Category Archives: Cybersecurity / 网络安全

Xiamen’s International Data Centre Measures: Cross-Border Lines, Data Classification and the Circuit Breaker

  • Five Xiamen municipal bodies issued the Trial Measures for the Administration of International Data Centres (Xia Shu Ju Gui [2026] No. 1) on 20 October 2026; they take effect on 25 November 2026 and remain valid for two years.
    厦门市五个市级部门于 2026 年 10 月 20 日印发《厦门市国际数据中心管理办法(试行)》(厦数据规〔2026〕1 号),自 2026 年 11 月 25 日起试行,有效期 2 年。
  • Only market entities registered and actually operating in Xiamen may build, operate or use the facility, and cross-border flow, regulatory audit and substantive compliance responsibility all rest with that local entity.
    只有登记注册并实际运营在厦门市的市场主体才能建设、运营或使用该设施,跨境数据流动、监管审计与合规主体责任均由该本地主体承担。
  • Connectivity must run through a basic telecommunications operator’s dedicated cross-border link and an international gateway approved by the Ministry of Industry and Information Technology; private networks, tunnelling and unlicensed lines are prohibited.
    跨境连通必须经由基础电信运营企业的跨境专用通信链路,并通过工业和信息化部批准的国际通信出入口局;私自架设虚拟专用网络、隧道技术或非法专线均被禁止。
  • Data must be classified and graded; exports of important data and of personal information above the assessment threshold still require a data export security assessment, and data subject to domestic storage rules may not leave over the link.
    数据须分类分级;重要数据出境以及达到申报门槛的个人信息出境仍须通过数据出境安全评估,要求境内存储的数据不得经该链路出境。
  • A circuit breaker lets the telecommunications authority cut the link on four defined triggers, with resumption only after rectification and approval.
    业务熔断机制允许电信主管部门在四类情形下切断链路,且仅在整改完成并经批准后方可恢复。
  • Eight business formats are cultivated, of which multinational internal management, overseas data processing, cross-border e-commerce and cross-border settlement map most directly onto existing foreign-invested operations.
    办法培育八类业态,其中跨国企业内部管理、来数加工、跨境电商与跨境结算与现有外资企业业务重合度最高。

Xiamen’s International Data Centre Measures: Cross-Border Lines, Data Classification and the Circuit Breaker | 厦门市国际数据中心管理办法(试行):跨境专线、数据分类与业务熔断

Overview

On 20 October 2026 five Xiamen municipal bodies issued the Trial Measures for the Administration of International Data Centres (Xia Shu Ju Gui [2026] No. 1): the Xiamen Municipal Data Administration, the Office of the Committee for Comprehensively Deepening Reform of the CPC Xiamen Municipal Committee, the Office of the Cyberspace Affairs Commission of the CPC Xiamen Municipal Committee, the Xiamen Municipal Communications Administration and the Xiamen Airport Economic Zone Headquarters. The measures take effect on 25 November 2026 and remain valid for two years.

They rest on five instruments: the State Council reply on piloting market-based allocation of production factors, the State Council General Office opinion on trade in services, Xiamen’s services-sector pilot plan, the Provisions on Promoting and Regulating Cross-Border Data Flows, and the Xiamen Special Economic Zone Data Regulations.

The question foreign companies ask is whether one dedicated line can join their operations inside and outside China. The measures answer it in three layers, and the answer is no at each unless the conditions are met separately.

What the measures establish

Article 2 fixes the perimeter. The measures apply to market entities registered and actually operating in Xiamen that build, operate or manage an international data centre, and to the whole of the international data business conducted on that infrastructure. Cross-border data flow, regulatory audit and the substantive compliance responsibility all sit with the locally registered market entity. A group therefore cannot hold the infrastructure in one entity, the traffic in a second and the liability in a third offshore.

Articles 4 and 5 define the two operative terms. An “international data centre” is a facility whose cross-border connectivity is provided by a basic telecommunications operator through a dedicated cross-border communications link, with network security isolation in place, connected internationally through an international communications gateway station approved by the Ministry of Industry and Information Technology, and which maintains a central equipment room and provides server hosting, cloud services and computing power services in a unified way. “International data business” means the collection, storage, processing and governance of overseas data using the international data line on that infrastructure.

Read together, the definitions close a door that had been left ajar: the centre is not a general-purpose exit.

Layer one: the infrastructure

Article 14 is the operational core. Construction and operation must meet national and industry technical standards, and the operator needs a telecommunications business operating licence and the associated qualifications. International data business must be accessed through the dedicated cross-border link provided by a basic telecommunications operator. Private connection to overseas nodes through virtual private networks, tunnelling technology or illegally operated international leased lines is expressly prohibited.

For a foreign group this removes the grey route entirely. An encrypted tunnel between a Shanghai office and a Singapore cloud region, or capacity bought from a reseller holding no international line licence, is not a tolerated practice that the measures now formalise. It is named as prohibited.

Layer two: the data

Article 15 sets the data-side conditions, and they are cumulative rather than alternative.

Data must be classified and graded according to national and industry standards. The export of important data, and the export of personal information that meets the threshold for a data export security assessment, must pass that assessment. Data that laws and regulations require to be stored within China must not flow directly abroad over the dedicated link. Cross-border transmission must use encryption technology and protocols meeting national cryptography management requirements.

The practical consequence is that the dedicated line changes the transport, not the permission. A company that would have needed a security assessment for a transfer before these measures came into force still needs one afterwards. The difference is that an assessment, once passed, can be implemented over a lawful link rather than an improvised one. The category of the data determines the route, and the route does not upgrade the category.

Layer three: liability and the circuit breaker

The measures adopt an allocation formula: whoever builds and operates is responsible, and whoever uses is responsible. Failure to honour a commitment, where it results in a data breach or another security incident or in other non-compliance, allows the regulator to suspend or terminate the relevant operations and to require rectification.

Alongside suspension sits a circuit breaker. Where monitoring identifies cross-border traffic that endangers national security, is suspected of ideological infiltration or of disseminating illegal and harmful information; where an important data breach or a large-scale unlawful outbound transfer of information has occurred or is seriously suspected of having occurred; where a sudden cyber attack or an exploited vulnerability may endanger overall network security; or where another emergency may cause serious consequences — the telecommunications authority issues a link-disposal instruction and the link is cut. Operations resume only after rectification is complete and the regulator has approved resumption.

A cut is an instruction to the link, not a court order and not a fine; it can occur while a commercial dispute is unresolved, and resumption is discretionary as to timing. Any process that depends on continuous cross-border flow should be modelled with an interruption scenario.

The business models the rules name

The measures list eight cultivated formats: data processing for overseas clients, cross-border e-commerce, cross-border settlement, gaming services, distance education, telemedicine, cross-border manufacturing and the internal management of multinational enterprises. Priority goes to international data services, computing power services and cloud services in supply chains and cross-border logistics, BRICS trade-in-services cooperation, cross-Strait integration, and artificial intelligence and large-model training.

For foreign investors, four of the eight map onto operations that already exist. Internal management is the most directly usable: a regional headquarters consolidating finance, human resources and supply-chain planning for Asia can place that workload on the infrastructure, provided the data is not in a category requiring domestic storage and any personal information crossing the threshold has cleared assessment.

Xiamen’s International Data Hub, a State Council-approved project, opened the hub and its international data service centre at the Emerging Science and Innovation Industrial Park in the Airport Economic Zone on 14 September 2026. That site is the physical referent for these rules. An operating coordination mechanism draws in the cyberspace affairs, data, communications administration, development and reform, industry and information technology, commerce, foreign affairs and public security departments, with the pilot park’s management body handling policy implementation, public services, factor guarantees, application scenarios, industry development and project attraction, and an international data service centre operating on a government-guided, park-managed, market-entity-operated basis.

Implications for foreign companies

Three checks, in order. First, is the contracting entity registered and actually operating in Xiamen? If the contract sits with a Hong Kong or Singapore entity, the measures do not apply and its cross-border flow is governed by the general regime. Second, is connectivity contracted with a basic telecommunications operator and terminated at an approved gateway? If not, the arrangement is on the prohibited list whatever its technical merits. Third, has the data been classified, and does any part require assessment or domestic storage? If yes, the line is available only for the portion that clears it.

The official reading is direct. The responsible official of the Xiamen Municipal Data Administration, as reported by Xiamen Daily, stated that an international data centre is not a channel that bypasses existing rules and lets all data leave freely, and that enterprises must still perform their compliance obligations according to data type, business scenario and the relevant law.

Sources

  • 厦门市数据管理局等五部门·关于印发《厦门市国际数据中心管理办法(试行)》的通知(厦数据规〔2026〕1号,2026-10-20印发),含办法全文:https://as.xm.gov.cn/zwgk/zfxx/zc/xzgfxwj/202610/t20261022_3022410.htm
  • 厦门市人民政府·政务公开(厦门市数据管理局政策文件发布栏目):https://www.xm.gov.cn/
  • 厦门网(厦门日报)·《厦门市国际数据中心管理办法(试行)》印发(2026-10-20),转载办法全文:https://www.xmnn.cn/news/xmxw/202610/t20261024_457606.html
  • 厦门网·厦门市国际数据枢纽港与金砖创新基地建设(2026-10-13):https://www.xmnn.cn/news/xmxw/202610/t20261013_455584.html

厦门市国际数据中心管理办法(试行):跨境专线、数据分类与业务熔断

概述

2026 年 10 月 20 日,厦门市数据管理局、中共厦门市委全面深化改革委员会办公室、中共厦门市委网络安全和信息化委员会办公室、厦门市通信管理局、厦门临空经济片区指挥部五个市级部门联合印发《厦门市国际数据中心管理办法(试行)》,文号为厦数据规〔2026〕1 号。办法自 2026 年 11 月 25 日起试行,有效期 2 年。

依据为五份文件:国务院关于部分地区要素市场化配置综合改革试点的批复、国务院办公厅关于以高水平开放推动服务贸易高质量发展的意见、《厦门市服务业扩大开放综合试点实施方案》《促进和规范数据跨境流动规定》以及《厦门经济特区数据条例》。

外资企业关心的问题是:能否用一条专线把境内外的业务打通。办法的答案分三层,每层的答案都是”不能”,除非各层条件分别满足。

办法确立了哪些规则

第二条划定适用范围:在厦门市登记注册并实际运营的市场主体,开展国际数据中心建设、运营、管理,以及依托该基础设施从事国际数据业务的全部活动。跨境数据流动、监管审计以及合规主体责任,均由厦门本地注册的市场主体承担。这意味着一个集团不能把基础设施放在一个主体、把流量放在第二个主体、把责任放在境外的第三个主体。

第四条与第五条界定两个核心概念。”国际数据中心”,指由基础电信运营企业提供跨境专用通信链路、落实网络安全隔离、经工业和信息化部批准的国际通信出入口局实现跨境连通,建有中心机房并统一提供服务器架设、云服务及算力服务的设施。”国际数据业务”,指依托国际数据中心、利用国际数据专线,对境外数据提供收集、存储、加工、治理等服务。

两条定义合起来读,等于把一扇原本半开的门关上了:这个中心不是一条通用出口。

第一层:基础设施

第十四条是操作层面的核心。建设运营须符合国家及行业技术标准,并需具备电信业务经营许可证及相关资质。国际数据业务必须通过基础电信运营企业提供的跨境专用通信链路接入。私自通过虚拟专用网络、隧道技术或非法经营的国际专线连接海外节点,被明令禁止。

对跨国集团而言,这一条把灰色路径整体移除了。在上海办公室与新加坡云区域之间自建加密隧道,或向并不持有国际专线牌照的转售商购买带宽,并非办法予以默许、随后加以规范的实践,而是被直接列为禁止事项。

第二层:数据

第十五条设定数据侧的条件,且这些条件累加适用,不能相互替代。

数据须按国家及行业标准分类分级。重要数据出境,以及符合应当申报数据出境安全评估条件的个人信息出境,必须通过数据出境安全评估。法律法规要求境内存储的数据,严禁通过跨境专用链路直接流向境外。跨境传输应采用符合国家密码管理要求的加密技术和协议。

实务上的结论是:专线改变的是传输方式,不是许可本身。在办法出台之前需要做安全评估的传输,在办法出台之后仍然需要。区别只在于,评估通过之后,可以落在一条合法链路上实施,而不是靠临时搭起来的通道。数据的类别决定走哪条路,路径不会反过来提升数据的类别。

第三层:责任与熔断

办法采用一条归责公式:谁建设运营谁负责,谁使用谁负责。未履行承诺造成数据泄露等安全事件或其他违规的,监管部门可暂停或终止相关运营活动,并要求整改。

与暂停并列的是业务熔断机制。监测到跨境流量存在危害国家安全、涉嫌意识形态渗透或传播违法有害信息;发生或重大涉嫌发生重要数据泄露、大规模信息非法出境;突发网络攻击或漏洞被恶意利用且可能危及整体网络安全;或其他可能引发严重后果的紧急状态——出现上述情形之一,由电信主管部门依法下达链路处置指令,执行熔断。整改完成并报经监管部门批准后,方可恢复。

熔断是对链路下达的指令,不是法院裁定,也不是罚款。它可能在一项商业争议尚未解决时发生,恢复时点由监管部门裁量。任何依赖持续跨境流动的流程,都应按中断场景建模。

办法点名的业态

办法列出八类培育业态:来数加工、跨境电商、跨境结算、游戏服务、远程教育、远程医疗、跨境制造、跨国企业内部管理。重点发展供应链与跨境物流、金砖国家服务贸易合作、两岸融合发展,以及人工智能与大模型训练等领域的国际数据服务、算力服务及云服务。

对外资企业而言,八类之中有四类与既有业务直接重合。跨国企业内部管理最易落地:一个负责亚洲财务、人力资源与供应链计划的区域总部,可以把这部分工作负载放到该基础设施上,前提是所涉数据不属于要求境内存储的类别,且跨境部分中的个人信息已通过评估。

厦门国际数据枢纽港为国务院批复项目,2026 年 9 月 14 日在厦门临空经济片区新兴科创产业园揭牌国际数据枢纽港暨国际数据服务中心,即上述规则对应的物理场所。运营协调机制由网信、数据、通信管理、发展改革、工业和信息化、商务、外办、公安等部门按职责分工推进,试点园区管理机构负责政策实施、公共服务、要素保障、场景应用、产业发展与项目招引,并建立”政府引导、园区管理、市场主体运营”的国际数据服务中心。

对外资企业的实务意义

按顺序做三项核查。第一,签约主体是否在厦门登记注册并实际运营?如果合同签在香港或新加坡主体名下,办法不适用,其跨境流动适用一般制度。第二,连通是否与基础电信运营企业签约、链路是否终止于获批出入口局?若不是,无论技术方案多么优越,该安排都在禁止清单上。第三,数据是否已完成分类分级,其中是否有部分需要评估或境内存储?若是,专线仅对通过的部分可用。

官方口径同样直接。厦门市数据管理局负责人在接受厦门日报采访时表示,国际数据中心并不是可以绕开现有规定、让所有数据自由出境的通道,企业仍须按数据类型、业务场景和相关法律履行合规义务。

来源

  • 厦门市数据管理局·《厦门市国际数据中心管理办法(试行)》印发通知(厦数据规〔2026〕1 号,2026-10-22):https://as.xm.gov.cn/zwgk/zfxx/zc/xzgfxwj/202610/t20261022_3022410.htm
  • 厦门网转厦门日报·办法解读,含”不是绕开规定的通道”表述(2026-10-24):https://www.xmnn.cn/news/xmxw/202610/t20261024_457606.html
  • 腾讯新闻·转载厦门日报解读(2026-10-24):https://news.qq.com/rain/a/20261024A0669900
  • 厦门网·金砖创新基地若干措施,含”深度参与建设厦门国际数据枢纽港”(2026-10-13):https://www.xmnn.cn/news/xmxw/202610/t20261013_455584.html

Digital Trade and IP Rules Go National: Reading the Data and Standards Chapters of China’s 77-Measure FTZ Rollout

  • Within the State Council’s replication of 77 pilot measures from the Shanghai Free Trade Zone, the digital trade and intellectual property blocks are the ones with the most direct bearing on foreign firms’ data and standards compliance.
    在国务院将上海自贸试验区 77 条试点措施复制推广的安排中,数字贸易与知识产权两个板块与外资企业的数据合规、标准合规关系最直接。
  • Under the high-standard digital trade rules block (15 measures), five measures — including drawing up a data export negative list — are replicated to other free trade zones, and ten measures — including implementing a data security management certification system — are replicated nationwide.
    率先实施高标准数字贸易规则板块共 15 条,其中制定数据出境负面清单等 5 条复制推广到其他自贸试验区,实施数据安全管理认证制度等 10 条复制推广到全国。
  • Under the intellectual property protection block (8 measures), all eight measures are replicated nationwide, including standardising the content of trademark registration announcements and preliminary approval announcements, and strengthening the protection of agrochemical test data.
    加强知识产权保护板块共 8 条,全部复制推广到全国,包括规范商标注册公告和初步审定公告内容、加强农用化学品试验数据保护等。
  • Within the services trade opening block (8 measures), four measures — including advancing the cross-border application of electronic payments — are replicated nationwide.
    加快服务贸易扩大开放板块共 8 条,其中推进电子支付跨境应用等 4 条复制推广到全国。
  • The China National Standardization Technical Committee for Cybersecurity released the Artificial Intelligence Security Governance Framework 3.0 in September 2026.
    全国网络安全标准化技术委员会于 2026 年 9 月发布《人工智能安全治理框架 3.0》。
  • SAMR sought public comment in September 2026 on the Measures for the Administration of Foreign-Language Versions of National Standards (Draft for Comment), providing that mandatory national standards, standards needed for import and export trade, and standards promoting products, technologies and services going global should be translated wherever possible; comments close on 10 October 2026.
    市场监管总局 2026 年 9 月就《国家标准外文版管理办法(征求意见稿)》公开征求意见,明确强制性国家标准、进出口贸易所需标准及促进产品技术服务”走出去”的国家标准应译尽译,意见反馈截止 2026 年 10 月 10 日。
  • On 10 September 2026 SAMR (the Standardization Administration of China) approved and issued 298 important national standards — 27 mandatory and 271 recommended — and on 20 September 2026 it said that during the 15th Five-Year Plan period at least 20,000 national standards would be developed or revised.
    2026 年 9 月 10 日市场监管总局(国家标准委)批准发布 298 项重要国家标准(强制性 27 项、推荐性 271 项);2026 年 9 月 20 日表示「十五五」时期将制修订国家标准不少于 2 万项。

Digital Trade and IP Rules Go National: Reading the Data and Standards Chapters of China’s 77-Measure FTZ Rollout | 数字贸易与知识产权规则走向全国:77 条自贸试验区试点措施中的数据与标准板块

Overview

The State Council has moved 77 mature, effective and risk-controllable pilot measures out of the Shanghai Free Trade Zone and into wider application. For foreign firms, the two blocks that matter most to day-to-day data and standards compliance are the high-standard digital trade rules and the intellectual property protection chapters, with a further relevant strand in the services trade opening chapter on electronic payments.

This article does not restate the full seven-part structure of the rollout. It concentrates on the digital trade and intellectual property blocks, and on the accompanying national standards and AI governance developments that give those blocks their operating content. Read together, they describe where China’s cross-border data and standards regime is being made uniform rather than experimental.

The digital trade block: 15 measures, split two ways

The high-standard digital trade rules block contains 15 measures, and its replication is split by destination:

  • five measures are replicated to other free trade zones, including drawing up a data export negative list; and
  • ten measures are replicated nationwide, including implementing a data security management certification system.

The split is instructive. The negative list, which is inherently a zone-level instrument tuned to local sectoral conditions, stays within the free trade zone family. The certification system, which is a national conformity assessment mechanism, is extended to the whole country. For a foreign group, the practical consequence is that the certification route becomes a nationwide compliance option, while negative-list relief remains a matter of which zone you sit in.

The quantitative thresholds that sit beneath these measures were set out by the Cyberspace Administration of China’s Questions and Answers on Policies and Regulations for Data Export Security Management (September 2026), released on 11 September 2026; that document, and the certification-versus-assessment boundary it draws, is dealt with separately and is not repeated here.

The intellectual property block: 8 measures, all nationwide

The intellectual property protection block contains 8 measures, and all eight are replicated nationwide. Two named examples show the direction of travel:

  • standardising the content of trademark registration announcements and preliminary approval announcements; and
  • strengthening the protection of agrochemical test data.

Both are transparency and data-protection measures rather than zone-specific concessions. Replicating them nationwide aligns trademark practice and test-data protection across the country, which matters to foreign holders of brands and of agrochemical registrations that rely on the consistency of Chinese procedures.

The electronic-payment strand in services trade

Within the services trade opening block of 8 measures, four measures are replicated nationwide, including advancing the cross-border application of electronic payments. For foreign firms in cross-border e-commerce and services, this is the strand that touches how cross-border settlement is conducted, and it moves with the same nationwide destination as the certification measures.

Standards: foreign-language versions and volume targets

The standards work runs alongside the digital trade measures. In September 2026, SAMR sought public comment on the Measures for the Administration of Foreign-Language Versions of National Standards (Draft for Comment). The draft provides that mandatory national standards, standards needed for import and export trade, and standards promoting products, technologies and services going global should be translated wherever possible (应译尽译). Comments close on 10 October 2026.

Two further data points set the scale. On 10 September 2026, SAMR (the Standardization Administration of China) approved and issued 298 important national standards, of which 27 are mandatory and 271 are recommended. On 20 September 2026, at a State Council Information Office press conference, Zhu Jianqiao, Director-General of SAMR’s Department of Network Transaction Supervision and Management, said that during the 15th Five-Year Plan period China would develop or revise at least 20,000 national standards, including at least 1,000 mandatory national standards, and would carry out at least 200 pilot applications of Chinese standards overseas.

For foreign firms, the foreign-language-version measures and the overseas pilot target point in one direction: Chinese standards are to be made more usable beyond the Chinese-language market, both by translation and by application abroad.

AI governance and enforcement context

On the AI side, the Artificial Intelligence Security Governance Framework 3.0 was released in September 2026 by the China National Standardization Technical Committee for Cybersecurity. It sits within the same standardisation ecosystem that the digital trade measures rely on.

On enforcement, the Cyberspace Administration of China published ten typical enforcement cases on 15 September 2026 in the fields of cybersecurity, data security and personal information protection. The cases cover website tampering, malicious programs, data breaches, personal information leaks, unlawful collection and use of personal information, unlawful export of personal information, failure to label AI-generated synthetic content, and new technologies or applications put into service without assessment.

What it means for foreign firms

Three practical points follow.

First, separate the national measures from the zone measures. The data security management certification system and the eight intellectual property measures are now nationwide, so they should be treated as the baseline for planning. The data export negative list remains a zone-level instrument, so its relief depends on where the entity is established.

Second, track the standards translation programme. The draft measures on foreign-language versions of national standards, with their “translate wherever possible” rule for mandatory standards and for import-export standards, will determine which standards are available in a language your teams can apply directly. The comment window closes on 10 October 2026.

Third, read the electronic-payment measure with your settlement arrangements. Advancing the cross-border application of electronic payments is now a nationwide measure, and it is the strand most likely to change how cross-border settlement is structured for e-commerce and services businesses.

Sources

  • 满洲里市人民政府 — 自贸试验区试点措施复制推广: https://www.manzhouli.gov.cn/News/show/1456699.html
  • 上海市互联网信息办公室 — 数字贸易与数据出境政策宣介: https://wsb.sh.gov.cn/ztzl/shzxzb/20260923/3a34c2e581c04aa4a3eb7dafedd9e404.html
  • 国家互联网信息办公室 — 执法典型案例(2026 年 9 月 15 日): https://www.cac.gov.cn/2026-09/15/c_1790876152357946.htm
  • 贵州省大数据发展管理局 — 《数据出境安全管理政策法规问答(2026 年 9 月)》: https://drc.guizhou.gov.cn/ztzl/wlaqzs/202609/t20260918_90894948.html
  • 工业和信息化部 — 标准工作信息: https://www.miit.gov.cn/jgsj/kjs/jscx/bzgf/art/2026/art_88f891ce12ee49eabf2a3585eca08ee1.html
  • 国家市场监督管理总局: https://www.samr.gov.cn/

Related reading

  • see also: Where data export certification ends and the security assessment begins (data-export-certification-thresholds)
  • see also: Replicating Shanghai’s high-standard trade rules across free trade zones (ftz-high-standard-rules-replication)

数字贸易与知识产权规则走向全国:77 条自贸试验区试点措施中的数据与标准板块

概述

国务院已将 77 条成熟、成效显著、风险可控的试点措施从上海自贸试验区向更大范围复制推广。对外资企业而言,与日常数据合规、标准合规关系最直接的是率先实施高标准数字贸易规则与加强知识产权保护两个板块,此外加快服务贸易扩大开放板块中关于电子支付跨境应用的内容亦具相关性。

本文不复述该轮复制推广的七方面全貌,而集中说明数字贸易与知识产权两个板块,以及为其提供运行内容的配套国家标准与人工智能治理进展。合起来看,它们描绘出中国的跨境数据与标准制度正从”试验”走向”统一”的方向。

数字贸易板块:15 条,按去向一分为二

率先实施高标准数字贸易规则板块共 15 条,其复制推广按去向分为两支:

  • 5 条复制推广到其他自贸试验区,其中包括制定数据出境负面清单;
  • 10 条复制推广到全国,其中包括实施数据安全管理认证制度。

这一分工颇具信息量。负面清单本质上是一种适配地方产业条件的试验区内工具,因而留在自贸试验区范围内;认证制度是一种全国性合格评定机制,因而扩展到全国。对外资集团而言,实际后果是:认证通道成为一项全国可用的合规选择,而负面清单的便利仍取决于企业落在哪个试验区。

这些措施之下的量化门槛,由国家互联网信息办公室《数据出境安全管理政策法规问答(2026 年 9 月)》给出,该问答于 2026 年 9 月 11 日发布;其中认证与安全评估之间的边界另有专文说明,此处不再重复。

知识产权板块:8 条,全部推广到全国

加强知识产权保护板块共 8 条,全部复制推广到全国。两个点名的例子显示其方向:

  • 规范商标注册公告和初步审定公告内容;
  • 加强农用化学品试验数据保护。

两者都属于透明度与数据保护类措施,而非试验区专属让利。将其推广到全国,使商标实务与试验数据保护在全国范围内趋于一致,这对依赖中国程序一致性的外国品牌持有人与农用化学品登记持有人尤为重要。

服务贸易中的电子支付内容

在加快服务贸易扩大开放板块的 8 条中,4 条复制推广到全国,其中包括推进电子支付跨境应用。对从事跨境电商与服务业务的外资企业而言,这是涉及跨境结算如何开展的一支,其去向与认证类措施同为全国。

标准:外文版管理与数量目标

标准工作与数字贸易措施并行推进。2026 年 9 月,市场监管总局就《国家标准外文版管理办法(征求意见稿)》公开征求意见。征求意见稿明确,强制性国家标准、进出口贸易所需标准及促进产品技术服务”走出去”的国家标准应译尽译。意见反馈截止 2026 年 10 月 10 日。

另有两个数据点显示规模。2026 年 9 月 10 日,市场监管总局(国家标准委)批准发布 298 项重要国家标准,其中强制性 27 项、推荐性 271 项。2026 年 9 月 20 日,在国务院新闻办公室发布会上,市场监管总局网络交易监督管理司司长朱剑桥表示,「十五五」时期将制修订国家标准不少于 2 万项、强制性国家标准不少于 1000 项,并开展不少于 200 个中国标准海外应用试点。

对外资企业而言,外文版管理办法与海外应用试点目标指向同一方向:通过翻译与境外应用,使中国标准在中文市场之外更可用。

人工智能治理与执法背景

在人工智能方面,《人工智能安全治理框架 3.0》由全国网络安全标准化技术委员会于 2026 年 9 月发布,与数字贸易措施所依托的标准生态同处一体。

在执法方面,国家互联网信息办公室于 2026 年 9 月 15 日发布网络安全、数据安全、个人信息保护等领域执法典型案例 10 例,涉及网页篡改、设置恶意程序、数据泄露、个人信息泄露、违法收集使用个人信息、违法出境个人信息、未落实人工智能生成合成内容标识要求、新技术新应用未经评估上线提供服务等情形。

对外资企业的实际含义

有三点实务结论。

其一,把全国性措施与试验区措施区分开。数据安全管理认证制度与 8 条知识产权措施现已覆盖全国,应作为规划基线;数据出境负面清单仍是试验区层面的工具,其便利取决于企业设立地。

其二,跟踪标准翻译计划。国家标准外文版征求意见稿对强制性标准与进出口所需标准采取”应译尽译”规则,将决定哪些标准能以团队可直接适用的语言获得。征求意见窗口于 2026 年 10 月 10 日关闭。

其三,把电子支付措施与自身结算安排对照阅读。推进电子支付跨境应用现已成为全国性措施,也是最可能改变电商与服务业务跨境结算架构的一支。

来源

  • 满洲里市人民政府 — 自贸试验区试点措施复制推广:https://www.manzhouli.gov.cn/News/show/1456699.html
  • 上海市互联网信息办公室 — 数字贸易与数据出境政策宣介:https://wsb.sh.gov.cn/ztzl/shzxzb/20260923/3a34c2e581c04aa4a3eb7dafedd9e404.html
  • 国家互联网信息办公室 — 执法典型案例(2026 年 9 月 15 日):https://www.cac.gov.cn/2026-09/15/c_1790876152357946.htm
  • 贵州省大数据发展管理局 — 《数据出境安全管理政策法规问答(2026 年 9 月)》:https://drc.guizhou.gov.cn/ztzl/wlaqzs/202609/t20260918_90894948.html
  • 工业和信息化部 — 标准工作信息:https://www.miit.gov.cn/jgsj/kjs/jscx/bzgf/art/2026/art_88f891ce12ee49eabf2a3585eca08ee1.html
  • 国家市场监督管理总局:https://www.samr.gov.cn/

相关阅读

  • 见:数据出境认证的量化边界与评估衔接(data-export-certification-thresholds)
  • 见:上海自贸试验区高标准经贸规则复制推广(ftz-high-standard-rules-replication)

China’s September 2026 Data Export Q&A: Where Certification Ends and the Security Assessment Begins

  • The Cyberspace Administration of China released its Questions and Answers on Policies and Regulations for Data Export Security Management (September 2026) on 11 September 2026.
    国家互联网信息办公室于 2026 年 9 月 11 日发布《数据出境安全管理政策法规问答(2026 年 9 月)》。
  • A non-critical-information-infrastructure operator that has cumulatively provided overseas, since 1 January of the year, personal information of 100,000 or more but fewer than one million individuals (excluding sensitive personal information), or sensitive personal information of fewer than 10,000 individuals, and that does not involve important data, may discharge its data export compliance duty through personal information export certification.
    非关键信息基础设施运营者自当年 1 月 1 日起累计向境外提供 10 万人以上、不满 100 万人个人信息(不含敏感个人信息),或者不满 1 万人敏感个人信息,且不包括重要数据的,可以通过个人信息出境认证方式履行数据出境合规义务。
  • Whatever the volume, a personal information processor that has an outbound transfer may apply to a professional certification institution for personal information export certification, to demonstrate that the processing activity meets GB/T 35273 and GB/T 46068.
    无论出境个人信息规模大小,存在个人信息出境行为的个人信息处理者均可向专业认证机构申请个人信息出境认证,证明有关处理活动符合 GB/T 35273 与 GB/T 46068 的要求。
  • Where certified processing later reaches one million or more individuals’ personal information (excluding sensitive personal information), or 10,000 or more individuals’ sensitive personal information, since 1 January of the year, a data export security assessment must be filed for through the provincial cyberspace administration to the national one.
    已通过认证但自当年 1 月 1 日起累计出境 100 万人以上个人信息(不含敏感个人信息),或者 1 万人以上敏感个人信息的,应当通过所在地省级网信部门向国家网信部门申报数据出境安全评估。
  • Quantity splitting and similar methods may not be used to route, through certification, personal information that should lawfully undergo a data export security assessment.
    不得采取数量拆分等手段,将依法应当通过数据出境安全评估的个人信息通过个人信息出境认证方式向境外提供。
  • Before applying for certification, a processor must perform its duties of notice, obtaining separate consent and conducting a personal information protection impact assessment, and may apply to any professional institution that has completed filing.
    申请认证前,个人信息处理者应当履行告知、取得个人单独同意、进行个人信息保护影响评估等义务,并可向已通过备案的任何一家专业机构申请。
  • By September 2026, nine free trade zones or ports — Tianjin, Beijing, Hainan, Shanghai, Zhejiang, Guangxi, Jiangsu, Chongqing and Fujian — had filed and published data export negative lists.
    截至 2026 年 9 月,全国已有天津、北京、海南、上海、浙江、广西、江苏、重庆、福建等 9 地自贸试验区(港)备案发布数据出境负面清单。

China’s September 2026 Data Export Q&A: Where Certification Ends and the Security Assessment Begins | 国家网信办《数据出境安全管理政策法规问答(2026 年 9 月)》:个人信息出境认证的量化边界与评估衔接

Overview

On 11 September 2026, the Cyberspace Administration of China released the Questions and Answers on Policies and Regulations for Data Export Security Management (September 2026). The stated purpose is to keep strengthening outreach on the policies and regulations governing data export security management, and to guide and help data processors to carry out data export activities efficiently and in compliance. Around 14 September 2026, local government websites carried the text in a concentrated way.

The Q&A answers three questions. Taken together, they fix the boundary between the personal information export certification route and the data export security assessment route, and they name the two national standards against which certification is measured. For foreign-invested groups whose China operations send personal information overseas, the document is the operative reference for deciding which compliance path applies.

Question 1: the certification route and its quantitative band

Personal information export certification is defined by reference to Article 38, paragraph 1, item 2 of the Personal Information Protection Law. It is a conformity assessment activity in which a professional certification institution that has lawfully obtained personal information protection certification qualifications certifies that a personal information processor’s activities — such as providing personal information overseas — comply with the relevant laws, administrative regulations, departmental rules, standards and technical specifications.

The Q&A then sets out the quantitative band in which certification is available. A non-critical-information-infrastructure operator may discharge its data export compliance duty through personal information export certification where, since 1 January of the year, it has cumulatively provided overseas either:

  • personal information of 100,000 or more but fewer than one million individuals (excluding sensitive personal information); or
  • sensitive personal information of fewer than 10,000 individuals;

and where important data is not involved.

The Q&A adds a broader proposition. Where a personal information processor has an outbound transfer of personal information, whatever the scale of that transfer, it may apply to a professional certification institution for personal information export certification, in order to demonstrate that the relevant processing activities meet the requirements of GB/T 35273 (Information Security Technology — Personal Information Security Specification) and GB/T 46068 (Data Security Technology — Security Certification Requirements for Personal Information Cross-border Processing Activities).

The Q&A also draws a hard line. Quantity splitting and similar methods may not be used to provide overseas, through the certification route, personal information that should lawfully pass through a data export security assessment.

Question 2: when certification must be upgraded to a security assessment

Certification is not a permanent alternative to assessment. Where a processor has already passed personal information export certification, but has cumulatively exported since 1 January of the year either one million or more individuals’ personal information (excluding sensitive personal information) or 10,000 or more individuals’ sensitive personal information, it must file for a data export security assessment with the national cyberspace administration through the provincial cyberspace administration at its place of domicile.

The Q&A provides a bridging mechanism for that filing. When filing, the processor may attach the relevant information about having passed personal information export certification, explaining the personal information protection level of the processor and of the overseas recipient, and the safeguards for personal information rights and interests. The cyberspace administration will take this into account in the course of the security assessment.

The practical reading is that certification and assessment sit on a single continuum. Certification is the lighter route in the lower band; when volume crosses the higher threshold, the assessment is mandatory, and the certification work already done is carried forward as supporting material rather than discarded.

Question 3: how to apply and what duties precede it

Applications follow the Measures for the Administration of Personal Information Export Certification. Before applying for certification to provide personal information overseas, a personal information processor must, in accordance with laws and administrative regulations, perform its duties of notice, obtaining separate consent, and conducting a personal information protection impact assessment.

On where to apply, the Q&A states that a processor may currently apply to any professional institution that has completed filing. By way of example it names the China Cybersecurity Review, Certification and Market Regulation Big Data Centre, with an enquiry telephone number of (010)88650936.

The two standards against which certification is measured

The Q&A ties certification to two national standards. GB/T 35273, the Information Security Technology — Personal Information Security Specification, is the long-standing baseline specification for personal information handling. GB/T 46068, the Data Security Technology — Security Certification Requirements for Personal Information Cross-border Processing Activities, is the certification-specific requirement for cross-border processing.

For a processor preparing an application, the two standards define the substance of the conformity assessment: not merely that a transfer was documented, but that the cross-border processing activity itself meets the stated requirements.

Enforcement context and regional negative lists

Two further developments frame the Q&A. On 15 September 2026, the Cyberspace Administration of China published ten typical enforcement cases in the fields of cybersecurity, data security and personal information protection. The cases cover website tampering, planting malicious programs, data breaches, personal information leaks, unlawful collection and use of personal information, unlawful export of personal information, failure to implement labelling requirements for AI-generated synthetic content, and new technologies or applications being put into service without assessment.

Separately, the Action Plan for Stabilising and Upgrading Foreign Investment calls for supporting free trade zones and national comprehensive pilot cities for opening up the services sector to explore, in more fields, scenario-based and field-level data export negative lists. By September 2026, nine free trade zones or ports had filed and published data export negative lists: Tianjin, Beijing, Hainan, Shanghai, Zhejiang, Guangxi, Jiangsu, Chongqing and Fujian.

What it means for foreign firms

Three practical points follow.

First, count before you choose a path. Determine whether the personal information you send overseas is sensitive, count the cumulative number of individuals since 1 January of the year, and test that figure against the bands. Below the certification thresholds, certification is available; at one million individuals’ non-sensitive personal information or 10,000 individuals’ sensitive personal information, the security assessment is mandatory.

Second, do not engineer around the thresholds. The Q&A expressly prohibits quantity splitting and similar methods. A structure that fragments flows to keep each stream under a threshold does not remove the underlying obligation, and the enforcement cases published on 15 September 2026 show that unlawful export of personal information is an active enforcement theme.

Third, use the certification work as assessment input. Where volume later crosses the higher band, the information about a passed certification may be attached to the assessment filing and will be taken into account. A processor that has documented its protection level and its safeguards is therefore better positioned when the heavier route is triggered.

Sources

  • 贵州省大数据发展管理局转载 — 《数据出境安全管理政策法规问答(2026 年 9 月)》: https://drc.guizhou.gov.cn/ztzl/wlaqzs/202609/t20260918_90894948.html
  • 国家互联网信息办公室 — 网络安全、数据安全、个人信息保护执法典型案例(2026 年 9 月 15 日): https://www.cac.gov.cn/2026-09/15/c_1790876152357946.htm
  • 上海市互联网信息办公室转载 — 数据出境政策宣介: https://wsb.sh.gov.cn/ztzl/shzxzb/20260923/3a34c2e581c04aa4a3eb7dafedd9e404.html
  • 商务部政策发布 — 《利用外资固稳促优行动方案》: https://www.mofcom.gov.cn/zcfb/index.html
  • 内蒙古自治区党委网信办 — 政策法规宣介: http://www.nmgwx.gov.cn/centralSpirit/19297.jhtml
  • 国家互联网信息办公室: https://www.cac.gov.cn/

Related reading

  • see also: Guangdong’s Greater Bay Area Data Special Zone Work Plan (gba-data-special-zone)
  • see also: Digital trade and IP rules in the 77-measure FTZ rollout (digital-trade-rules-replication)

国家网信办《数据出境安全管理政策法规问答(2026 年 9 月)》:个人信息出境认证的量化边界与评估衔接

概述

2026 年 9 月 11 日,国家互联网信息办公室发布《数据出境安全管理政策法规问答(2026 年 9 月)》。其目的表述为:持续加强数据出境安全管理政策法规宣介,指导和帮助数据处理者高效合规开展数据出境活动。2026 年 9 月 14 日前后,各地政府网站对该问答进行了集中转载。

该问答回答三个问题。三者合起来,划定了个人信息出境认证通道与数据出境安全评估通道之间的边界,并点明了衡量认证的两项国家标准。对在华业务需要向境外提供个人信息的外资集团而言,这份问答是判断适用哪条合规路径的现行依据。

问一:认证通道的适用情形与量化区间

个人信息出境认证以《个人信息保护法》第三十八条第一款第二项为规范依据。它是指由依法取得个人信息保护认证资质的专业认证机构,证明个人信息处理者向境外提供个人信息等个人信息处理活动符合相关法律、行政法规、部门规章、标准、技术规范的合格评定活动。

问答随后给出认证通道可用的量化区间。非关键信息基础设施运营者自当年 1 月 1 日起累计向境外提供下列情形之一的,可以通过个人信息出境认证方式履行数据出境合规义务:

  • 10 万人以上、不满 100 万人个人信息(不含敏感个人信息);或者
  • 不满 1 万人敏感个人信息;

且不包括重要数据。

问答还给出一个更宽的口径:个人信息处理者存在个人信息出境行为,不论出境个人信息规模大小,均可向专业认证机构申请个人信息出境认证,以证明有关处理活动符合 GB/T 35273《信息安全技术 个人信息安全规范》、GB/T 46068《数据安全技术 个人信息跨境处理活动安全认证要求》的要求。

问答同时划出一条硬线:不得采取数量拆分等手段,将依法应当通过数据出境安全评估的个人信息通过个人信息出境认证方式向境外提供。

问二:何时认证必须升级为安全评估

认证并非安全评估的永久替代。已通过个人信息出境认证,但自当年 1 月 1 日起累计出境100 万人以上个人信息(不含敏感个人信息)或者1 万人以上敏感个人信息的,个人信息处理者应当通过所在地省级网信部门向国家网信部门申报数据出境安全评估。

问答为这次申报设置了衔接机制。申报时,可以将通过个人信息出境认证的有关情况附后,说明个人信息处理者和境外接收方的个人信息保护水平、个人信息权益保障等情况,网信部门将在安全评估过程中予以参考。

实操理解是:认证与评估位于同一条连续线上。较低区间以认证为较轻路径;当数量跨过较高门槛,评估即为强制,此前完成的认证工作作为支撑材料承接下来,而非作废。

问三:如何申请,以及申请前应履行的义务

申请按照《个人信息出境认证办法》办理。个人信息处理者在申请认证向境外提供个人信息前,应当按照法律、行政法规的规定履行告知、取得个人单独同意、进行个人信息保护影响评估等义务。

关于向谁申请,问答指出,目前可向已通过备案的任何一家专业机构申请;文中举例机构为中国网络安全审查认证和市场监管大数据中心,咨询电话为 (010)88650936。

衡量认证的两项国家标准

问答把认证与两项国家标准绑定。GB/T 35273《信息安全技术 个人信息安全规范》是个人信息处理的长期基础性规范;GB/T 46068《数据安全技术 个人信息跨境处理活动安全认证要求》则是面向跨境处理活动的专门认证要求。

对准备申请的处理者而言,这两项标准界定了合格评定的实质:不仅是出境行为被记录,而是跨境处理活动本身符合规定要求。

执法背景与区域性负面清单

另有两项进展为问答提供背景。2026 年 9 月 15 日,国家互联网信息办公室发布网络安全、数据安全、个人信息保护等领域执法典型案例 10 例,涉及网页篡改、设置恶意程序、数据泄露、个人信息泄露、违法收集使用个人信息、违法出境个人信息、未落实人工智能生成合成内容标识要求、新技术新应用未经评估上线提供服务等情形。

此外,《利用外资固稳促优行动方案》提出,支持自由贸易试验区、国家服务业扩大开放试点城市探索在更多领域制定场景化、字段级数据出境负面清单。截至 2026 年 9 月,全国已有天津、北京、海南、上海、浙江、广西、江苏、重庆、福建等 9 地自贸试验区(港)备案发布数据出境负面清单。

对外资企业的实际含义

有三点实务结论。

其一,先盘清数量,再选择路径。判断拟出境个人信息是否属于敏感个人信息,统计自当年 1 月 1 日起的累计人数,并对照量化区间:低于认证门槛可走认证;达到 100 万人个人信息(不含敏感个人信息)或 1 万人敏感个人信息,则安全评估为强制。

其二,不要围绕门槛做技术性安排。问答明确禁止数量拆分等手段。把数据流切碎以使每一支低于门槛的架构,并不能消除底层义务;2026 年 9 月 15 日发布的执法典型案例也显示,违法出境个人信息是当前的执法重点之一。

其三,把认证成果用作评估输入。当数量日后跨过较高区间时,通过认证的有关情况可以附于评估申报材料并予以参考。已经记录自身保护水平与保障措施的处理者,在触发较重路径时因此处于更有利的位置。

来源

  • 贵州省大数据发展管理局转载 — 《数据出境安全管理政策法规问答(2026 年 9 月)》:https://drc.guizhou.gov.cn/ztzl/wlaqzs/202609/t20260918_90894948.html
  • 国家互联网信息办公室 — 网络安全、数据安全、个人信息保护执法典型案例(2026 年 9 月 15 日):https://www.cac.gov.cn/2026-09/15/c_1790876152357946.htm
  • 上海市互联网信息办公室转载 — 数据出境政策宣介:https://wsb.sh.gov.cn/ztzl/shzxzb/20260923/3a34c2e581c04aa4a3eb7dafedd9e404.html
  • 商务部政策发布 — 《利用外资固稳促优行动方案》:https://www.mofcom.gov.cn/zcfb/index.html
  • 内蒙古自治区党委网信办 — 政策法规宣介:http://www.nmgwx.gov.cn/centralSpirit/19297.jhtml
  • 国家互联网信息办公室:https://www.cac.gov.cn/

相关阅读

  • 见:广东大湾区数据特区工作方案(gba-data-special-zone)
  • 见:77 条试点措施中的数字贸易与知识产权规则(digital-trade-rules-replication)

China’s public security cyberspace inspection regime gets a major upgrade from 1 October 2026

  • The Ministry of Public Security issued Order No. 176, the Measures for Public Security Organs’ Supervision and Inspection of Cyberspace Security, on 6 August 2026; the rules take effect on 1 October 2026 and repeal the 2018 Internet security inspection rules (Order No. 151).
    2026年8月6日,公安部发布《公安机关网络空间安全监督检查办法》(公安部令第176号),自2026年10月1日起施行,2018年发布的《公安机关互联网安全监督检查规定》(公安部令第151号)同时废止。
  • The new measures shift the scope from “internet security” to “cyberspace security”, defined as the combination of network security, data security and information security, and list eight categories of inspection targets including data processors and personal-information processors.
    新办法把监管口径从”互联网安全”升级为”网络空间安全”(即网络安全、数据安全、信息安全三位一体),并列出八类检查对象,数据处理者、个人信息处理者被明确纳入。
  • Routine on-site inspection of operators of Level-3-and-above classified-protection networks and critical information infrastructure operators is set at once a year, and results of inspections already conducted by other competent authorities in the same year are reused to avoid duplicate checks.
    对网络安全等级保护第三级(含)以上网络运营者、关键信息基础设施运营者,每年开展一次日常性现场检查;本年度其他主管部门已开展的现场检查,公安机关复用检查结果、不再重复检查。
  • Algorithm security is a newly added inspection item: public security organs will check whether entities have implemented their primary responsibility for algorithm safety, including recommender-system management institutions and technical safeguards.
    算法安全是新纳入的重点检查内容:公安机关将核查被检查对象是否依法落实算法安全主体责任、建立健全算法推荐管理制度和技术措施。
  • Inspection methods now include online patrol, vulnerability scanning and, for city-level-and-above organs, remote penetration testing with three working days’ advance notice; technical support may be outsourced to qualified security service providers under police command.
    检查方式升级为线上巡查、漏洞扫描,设区的市级以上公安机关可开展漏洞探测、渗透性测试等远程检测(提前三个工作日告知),并可委托具备能力的网络安全服务机构在民警指挥下提供技术支持。
  • For enterprises, the message is that data security and personal-information protection duties are now subject to direct routine supervision by public security organs, alongside traditional duties such as record-filing of networked units, log retention and classified-protection compliance.
    对企业的直接含义是:联网备案、日志留存、等级保护等传统义务之外,数据安全与个人信息保护义务也将接受公安机关常态化直接监管,合规准备应提前完成。

China’s public security cyberspace inspection regime gets a major upgrade from 1 October 2026 | 公安机关网络空间安全监督检查新规:2026年10月1日起全面升级

Overview

On 6 August 2026, the Ministry of Public Security (MPS) published the Measures for Public Security Organs’ Supervision and Inspection of Cyberspace Security (公安机关网络空间安全监督检查办法) as Order No. 176, adopted at the second MPS ministerial meeting on 1 July 2026. The Measures take effect on 1 October 2026, at which point the 2018 Provisions on Public Security Organs’ Inspection of Internet Security (Order No. 151) are repealed. For any organisation operating online in China — including the China subsidiaries of foreign enterprises, data processors and app or platform operators — the change is more than a renaming exercise: it redefines what public security organs may inspect, whom they may inspect, and how often.

What changed: from “internet security” to “cyberspace security”

The most fundamental change is conceptual. The old 2018 rules governed “internet security” in the narrow sense of protecting externally facing online business. Order No. 176 defines cyberspace security as the sum of network security, data security and information security, and states that the measures apply to supervision and inspection of whether network operators, data processors, personal-information processors and others have performed their statutory security obligations.

The practical effect is that internal systems, business networks that are not open to the public, data warehouses and employee or customer personal-information repositories all fall within the inspection remit. Being “not an internet company” is no longer a defence: the Measures explicitly cover data processing and personal-information processing activities wherever they occur.

Eight categories of inspection targets

Article 6 lists eight categories of entities that public security organs may inspect according to operational need:

  1. Internet service providers — access, data centres, content delivery, domain services and information services;
  2. Public internet access providers;
  3. Network operators and their constructors and maintainers;
  4. Critical information infrastructure (CII) operators and their constructors and maintainers;
  5. Providers of network products and services;
  6. Data processors;
  7. Personal-information processors;
  8. Other entities lawfully subject to inspection.

Entities that have previously suffered a network or data security incident, or that were administratively penalised for failing to perform statutory security obligations without rectifying as required, are flagged for key-focus inspection. Foreign-invested companies that operate cloud, data, e-commerce or software businesses in China should assume they fall within categories 5 to 7.

What is inspected: eleven core items plus special-event items

Article 7 sets out eleven focal points for routine inspection, including whether the entity has:

  • completed networked-unit record-filing and reported access-provider and user information and changes;
  • established and implemented network, data and information security management institutions and operating procedures;
  • recorded and retained user registration information and internet activity logs as required;
  • performed classified-protection (MLPS) obligations — level-setting filing, level testing, build-and-rectify and self-inspection;
  • performed CII security-protection obligations;
  • adopted technical measures against computer viruses, cyber attacks and intrusions;
  • rectified security vulnerabilities and eliminated risks;
  • taken preventive measures against content that laws and regulations prohibit;
  • implemented its primary responsibility for algorithm security, including recommender-system management institutions and technical measures (new);
  • performed data-security and personal-information-protection obligations (now enforceable directly by public security organs);
  • provided technical support and assistance to public security organs for national security, counter-terrorism and crime investigation.

During major national security-protection task periods (for example, major summits or events), Article 8 adds special-purpose inspection items covering security plans, risk assessments, emergency-response plans and drills, and incident reporting.

How inspection happens: layered methods with burden-reduction safeguards

The Measures deliberately combine stronger methods with burden-reduction for businesses:

  • Online patrol (Article 4): information-patrol, information-audit capability testing and vulnerability scanning conducted in ways that do not disrupt normal operations. Information-audit capability tests require three working days’ advance notice.
  • Remote detection: city-level-and-above public security organs may carry out vulnerability probing and penetration testing on network facilities and systems other than CII, again with three working days’ notice, without interference, and with coordination notices to cyberspace-administration and industry authorities.
  • On-site inspection (Articles 5, 11–12): conducted by county-level-and-above organs at the operator’s main place of operations; at least two police officers with police ID and an inspection notice; measures include entering premises, questioning responsible persons, reviewing and copying relevant information, and viewing security-protection measures.
  • Frequency cap (Article 9): for operators of Level-3-and-above MLPS networks and CII operators, routine on-site inspection is limited to once a year; if another competent authority has already conducted an on-site inspection in the same year, the public security organ reuses those results rather than repeating the check.
  • Inter-agency coordination (Articles 9–10): for daily inspections in sectors with their own regulators (telecom, energy, transport, water, finance, defence science and industry), the relevant regulators and cyberspace-administration bodies are informed five working days in advance, and joint inspections are preferred where proposed. Nationwide multi-sector campaigns require approval from the Central Cyberspace Affairs Commission.
  • Outsourced technical support (Article 13): public security organs may commission qualified security service providers, who must act under police command, sign confidentiality commitments and be subject to background vetting where penetration testing is involved.

Outcomes and enforcement

Article 14 forbids charging fees or requiring the purchase of designated products. Where risk hidden dangers are found, organs direct rectification (Article 16). Where problems do not yet constitute illegal activity, the organ may issue a public-security reminder letter, send a letter to the competent industry authority, or publish a public notice without naming the entity (Article 17). Legal representatives or principal responsible persons of operators may be summoned for talks where significant risks or incidents exist (Article 19). Breaches of statutory duties are pursued under the Cybersecurity Law, the Data Security Law, the Personal Information Protection Law, the CII Security Protection Regulations, the Network Data Security Administration Regulations and related rules (Article 16). Information obtained during inspection may be used only for cyberspace-security purposes and must be kept confidential (Article 20).

Practical steps for foreign-invested companies

  • Treat 1 October 2026 as a compliance deadline. Run an internal gap assessment before the Measures take effect, covering MLPS level-setting and testing validity, log retention, vulnerability management and the completeness of security-management documentation.
  • Extend the compliance scope to data and personal information. Document data classification and grading, important-data protection measures, risk assessments and personal-information-protection impact assessments, because these duties are now directly inspectable by public security organs rather than being treated as cyber-office-only matters.
  • Review algorithm practices. If the business operates content recommendation, search or ranking algorithms, confirm that algorithm primary-responsibility institutions and recommender-management technical safeguards are in place and properly documented.
  • Prepare an inspection kit. Designate a responsible manager and point of contact for inspections; keep complete records of record-filing certificates, testing reports, rectification logs, incident and drill records.
  • Leverage the coordination safeguards. If another regulator (e.g. the Cyberspace Administration of China or an industry authority) has already inspected the entity this year, be ready to cite that inspection to avoid duplicate on-site checks.
  • If a security firm is engaged by the authority, co-operate under the rules: technical supporters act under police command and must sign confidentiality commitments covering trade secrets and personal information.

Sources

  • 中国网(据公安部网站消息)— 网安监督检查新规明确检查内容等 10月1日起施行: https://news.china.com.cn/2026-08/07/content_118638649.shtml
  • 安全内参 — 《公安机关网络空间安全监督检查办法》出台,企业合规面临哪些新变化?: https://www.secrss.com/articles/92931
  • 微博专栏(2026年8月国内网络安全领域重要政策及标准速览,收录176号令要点): https://weibo.com/ttarticle/p/show?id=2309405339428934779056

Related reading

  • see also: how MLPS and CII obligations apply to foreign-invested enterprises (37-cybersecurity-mlps-fie)

公安机关网络空间安全监督检查新规:2026年10月1日起全面升级

概述

2026年8月6日,公安部公布《公安机关网络空间安全监督检查办法》(公安部令第176号)。该办法已于2026年7月1日经公安部第2次部务会议审议通过,自2026年10月1日起施行,2018年发布的《公安机关互联网安全监督检查规定》(公安部令第151号)同时废止。对任何在中国境内开展线上业务的组织——包括外资企业在华子公司、数据处理者、APP与平台运营者——这次调整不是简单的改名:它重新定义了公安机关可以检查什么、检查谁、多久查一次。

变化核心:从”互联网安全”到”网络空间安全”

最根本的变化在理念层面。旧版2018年规则治理的是狭义”互联网安全”,即对外在线业务防护。176号令把网络空间安全定义为网络安全、数据安全、信息安全三者的总和,并明确本办法适用于对网络运营者、数据处理者、个人信息处理者等履行法定安全义务情况开展的监督检查。

实际效果是:内网系统、不对公开放网的业务系统、数据仓库、员工或客户个人信息库,全部落入监管视野。”我们不是互联网公司”不再是挡箭牌——办法明确覆盖一切数据与个人信息处理活动,无论其发生在何处。

八类检查对象

第六条规定,公安机关可根据维护网络空间安全需要,对以下对象依法开展监督检查:

  1. 提供互联网接入、数据中心、内容分发、域名服务、信息服务等的互联网服务提供者;
  2. 公共上网服务提供者;
  3. 网络运营者及其建设者、维护者;
  4. 关键信息基础设施运营者及其建设者、维护者;
  5. 网络产品、服务的提供者;
  6. 数据处理者;
  7. 个人信息处理者;
  8. 其他依法可以监督检查的对象。

曾发生网络安全、数据安全等事件,或曾因未履行法定义务被行政处罚且未按要求整改的,列为重点监督检查对象。在华从事云、数据、电商或软件业务的外资企业,应默认自身落在第5至第7类范围内。

查什么:十一项重点内容加重大活动专项

第七条列出日常检查十一项重点,包括是否:依法办理联网单位备案并报送接入单位和用户信息及变更情况;依法制定并落实网络、数据、信息安全管理制度和操作规程;依法记录并留存用户注册信息和上网日志;依法履行等级保护(定级备案、等级测评、建设整改、自查);依法履行关键信息基础设施安全保护义务;依法采取防范病毒、攻击、入侵的技术措施;针对漏洞隐患采取整改措施;依法对法律法规禁止发布或传输的信息采取防范措施;依法落实算法安全主体责任、建立健全算法推荐管理制度和技术措施(新增);依法履行数据安全与个人信息保护义务(现由公安机关直接监管);依法为国家安全、反恐、侦查犯罪提供技术支持与协助。

在国家重大安全保卫任务期间,第八条增加专项检查内容,覆盖安保工作方案与责任分工、风险评估与风险管控、应急预案与演练、应急设施有效性及事件报告处置情况。

怎么查:多元手段叠加”减负”机制

办法有意把更强的手段与为企业减负的制度结合起来:

  • 线上巡查(第四条):网络信息巡查、信息审核能力测试、漏洞扫描等不影响正常业务运行的方式。开展信息审核能力测试须提前三个工作日告知。
  • 远程检测:设区的市级以上公安机关可通过漏洞探测、渗透性测试,对关键信息基础设施以外的网络设施和信息系统进行远程检测,同样提前三个工作日告知,不得干扰破坏,并通报同级网信与行业主管部门。
  • 现场检查(第五、十一、十二条):由运营机构所在地县级以上公安机关实施;民警不少于二人并出示警官证与监督检查通知书;措施包括进入场所、问询负责人、查阅复制信息、查看安全保护技术措施运行情况、开展技术检测。
  • 频次上限(第九条):对等保三级(含)以上网络运营者与关基运营者,日常性现场检查每年一次;本年度其他主管部门已现场检查的,公安机关复用结果、不再重复检查。
  • 跨部门协同(第九、十条):对电信、能源、交通、水利、金融、国防科技工业等行业开展日常性现场检查,应提前五个工作日告知网信与行业主管部门,对方提出联合检查的应联合开展;公安部部署跨行业跨部门检查须报中央网络安全和信息化委员会审批。
  • 委托技术支持(第十三条):公安机关可委托具备能力的网络安全服务机构或专门人员提供技术支持,受托方须在民警指挥下开展工作、签署保密承诺书;涉及渗透测试的机构及人员须经背景审查并全流程安全管理。

结果运用与法律责任

第十四条禁止向被检查对象收取费用或要求购买指定产品服务。发现风险隐患的,督促指导整改(第十六条)。问题尚不构成犯罪的,可发放公安提示函、向行业主管部门发函或向社会发布不点名通告(第十七条)。存在较大风险或发生安全事件的,可对法定代表人、主要负责人进行约谈(第十九条)。未履行法定义务的,依据《网络安全法》《数据安全法》《个人信息保护法》《关键信息基础设施安全保护条例》《网络数据安全管理条例》等追究法律责任(第十六条)。检查获取的信息只能用于维护网络空间安全之目的,并须保密(第二十条)。

外资企业实务建议

  • 把2026年10月1日当作合规时点。 在办法施行前完成内部差距评估,覆盖等保定级与测评有效性、日志留存、漏洞闭环整改、安全管理制度文档完整性。
  • 把合规范围扩展到数据与个人信息。 数据分类分级、重要数据保护措施、风险评估、个人信息保护影响评估等文档要备齐,因为上述义务现在由公安机关直接检查,不再是”网信部门的事”。
  • 审查算法实践。 若经营内容推荐、搜索或排序算法业务,确认算法安全主体责任机构与推荐管理技术措施到位并可举证。
  • 准备迎检资料包。 指定责任人与迎检对接人;备案证书、测评报告、整改记录、事件与演练记录等完整留存。
  • 善用协同减负机制。 若本年度网信部门或行业主管部门已检查过,准备引用检查结论以避免重复现场检查。
  • 配合技术支持机构时注意边界。 受托机构在民警指挥下工作并签保密承诺书,涉商业秘密与个人信息事项须谨慎配合。

来源

  • 中国网(据公安部网站消息)— 网安监督检查新规明确检查内容等 10月1日起施行: https://news.china.com.cn/2026-08/07/content_118638649.shtml
  • 安全内参 — 《公安机关网络空间安全监督检查办法》出台,企业合规面临哪些新变化?: https://www.secrss.com/articles/92931
  • 微博专栏(2026年8月国内网络安全领域重要政策及标准速览,收录176号令要点): https://weibo.com/ttarticle/p/show?id=2309405339428934779056

相关阅读

  • 见:外资企业的等保与关基义务如何适用(37-cybersecurity-mlps-fie)

Cybersecurity & MLPS (等保) for FIEs

  • China operates a mandatory, tiered cybersecurity regime (MLPS + CII) that FIEs cannot opt out of; MLPS is the baseline obligation for virtually every network operator.
    中国实行强制、分层的网络安全制度(等保+关键信息基础设施),外商投资企业无法退出;等保几乎适用于所有网络运营者,是合规底线。
  • MLPS 2.0 (GB/T 22239-2019, effective 1 December 2019) has five steps: grading, filing, remediation, graded assessment, and supervision.
    等保2.0(GB/T 22239-2019,2019年12月1日施行)含定级、备案、建设整改、等级测评、监督检查五步骤。
  • Systems are graded into five levels (L1–L5); most foreign commercial operations land at Level 2 or 3, with Level 3 requiring an annual assessment.
    系统分为五级(第一至第五级);多数外资商业运营落在二级或三级,三级须每年测评。
  • CII operators (telecom, finance, energy, transport, etc.) must localize important data and personal information, pass a CAC cross-border security assessment, and meet duties beyond MLPS.
    关键信息基础设施运营者(电信、金融、能源、交通等)须境内存储重要数据和个人信息、通过网信办跨境安全评估,并承担对等保之外的额外义务。
  • The China subsidiary (not the overseas HQ) is normally the MLPS filing party; material changes such as new data types or M&A can reset the grading.
    备案主体通常为境内子公司(非境外总部);新数据类型、并购等重大变更可重置系统定级。
  • Enforcement is active—non-compliance can block a product launch, trigger fines, and intersect with PIPL and the Data Security Law.
    执法真实且日趋活跃——不合规可阻断产品上线、招致罚款,并与《个人信息保护法》及《数据安全法》相交叠。

Cybersecurity & MLPS (等保) for FIEs | 网络安全与等保(外资适用)

Overview

China operates a tiered, legally mandated cybersecurity regime that foreign-invested enterprises (FIEs) cannot opt out of. Two concepts sit at its centre: the Multi-Level Protection Scheme (MLPS, 网络安全等级保护, often called “等保”) and Critical Information Infrastructure (CII, 关键信息基础设施). MLPS is the baseline obligation for virtually every network operator in China; CII is a higher, security-hardened subset applied to systems whose disruption would threaten national security, the economy, or public welfare. Both rest on the Cybersecurity Law (effective 1 June 2017) and are administered by the public-security authorities (MLPS) and the cyberspace regulator (CII identification).

This article explains how MLPS 2.0 works, what the five grading levels mean for an FIE, how CII is designated and what extra duties it triggers, and the practical steps a foreign company should take to stay compliant.

MLPS 2.0: the graded-protection baseline

MLPS 2.0 is the current generation of China’s graded cybersecurity protection system, anchored by the national standard GB/T 22239-2019 (“Baseline for Graded Cybersecurity Protection”), released in May 2019 and effective 1 December 2019. It replaced the 1.0 framework and expanded coverage from traditional IT systems to cloud computing, mobile internet, IoT, industrial control systems, and big data.

The scheme has five steps: (1) grading (定级) — the operator assesses the impact of a breach and assigns a level; (2) filing/registration (备案) — the grading is submitted to and recorded by the public-security bureau; (3) construction and remediation (建设整改) — the system is brought up to the standard’s technical and management controls; (4) graded assessment (等级测评) — an accredited third-party evaluator tests compliance; and (5) supervision and review (监督检查) — ongoing oversight by the public-security authority.

Under the Cybersecurity Law, network operators that fail to implement MLPS face rectification orders, fines, and — for serious cases — penalties against responsible individuals. For an FIE, MLPS compliance is typically a precondition for launching customer-facing systems, passing partner security reviews, and obtaining the data-export and licensing approvals described in related articles.

The five protection levels

MLPS divides systems into five levels by the harm that a breach would cause:

  • Level 1 (自主保护级): damage limited to the legitimate interests of citizens, legal persons, or other organizations, with no harm to the state, social order, or public interest. Often self-assessed.
  • Level 2 (指导保护级): serious damage to those interests, or harm to social order/public interest, but no harm to national security. Requires filing; generally assessed periodically (commonly every two years).
  • Level 3 (监督保护级): serious harm to social order/public interest, or harm to national security. The most common level for commercial systems of meaningful scale; requires filing and an annual graded assessment.
  • Level 4 (强制保护级): particularly serious harm to social order/public interest, or serious harm to national security — typical of core finance, energy, and transport systems; assessed at least every six months.
  • Level 5 (专控保护级): particularly serious harm to national security — state-controlled, rarely seen in the private sector.

Most foreign-invested commercial operations land at Level 2 or Level 3. A practical rule: if the system handles large numbers of users, payment data, or personal information, expect Level 3 and an annual assessment obligation.

Critical Information Infrastructure (CII)

Beyond MLPS, the Cybersecurity Law requires the state to identify Critical Information Infrastructure — networks and systems in key industries (telecom, energy, transport, water, finance, public services, e-government, and national defence) whose impairment or data leakage would cause serious harm to national security, the national economy, people’s livelihoods, or the public interest. CII is designated by the relevant industry regulator and coordinated by the CAC.

CII operators owe duties that go well beyond MLPS: they must (a) store important data and personal information collected/generated in China on domestic soil; (b) pass a CAC security assessment before any cross-border transfer (see the PIPL article); (c) conduct regular security testing and risk assessment; (d) engage only vendors and products that meet CII security requirements; and (e) comply with dedicated CII security-protection regulations (the State Council’s 2021 CII Security Protection Regulation). An FIE in a sensitive sector — a cloud operator, a payment institution, a large platform, or a utility-linked business — should assume it may be designated CII and plan accordingly.

Obligations and enforcement for foreign-invested enterprises

For an FIE, the key compliance actions are: confirm which Chinese entity owns each system (the in-China subsidiary, not the overseas parent, is usually the filing party); grade and file every internet-facing or data-bearing system with the local public-security bureau; remediate to GB/T 22239-2019; and commission an accredited assessor. Level 3+ systems need recurring annual assessments, and any material change (new data type, new region, M&A) can reset the grading.

Enforcement is real and increasingly active. Public-security authorities conduct spot checks, and regulators share findings across the MLPS–CII–data-export stack. Non-compliance can block a product launch, trigger fines, and — where personal information or important data is involved — intersect with PIPL and the Data Security Law. Foreign groups should treat MLPS as a continuing programme owned by a local accountable executive, not a one-time certificate.

What to do next

  • Inventory your systems in China and assign a business owner to each; the China subsidiary (not the HQ) is normally the MLPS filing party.
  • Grade every system honestly against GB/T 22239-2019; expect Level 2–3 for user-facing or data-bearing systems.
  • File and remediate with an accredited evaluator, and schedule the recurring assessment (annual for Level 3).
  • Assess CII exposure if you operate in telecom, finance, energy, transport, or platform services; prepare for domestic storage and CAC assessment.
  • Build a running compliance programme (policies, logs, drills) rather than a one-off audit, and coordinate with your PIPL and data-export obligations.

Sources

Related reading

  • see also: PIPL compliance for foreign companies — CII and threshold processors must localize and pass CAC assessment before export.
  • see also: Anti-monopoly & merger control filing — M&A can reset system grading and trigger new filings.
  • see also: VIE structures & regulatory trends — data-security review of offshore listings.

网络安全与等保(外资适用)

概述

中国建立了分层、法定的网络安全制度,外商投资企业(FIE)无法选择退出。其两大核心概念是网络安全等级保护制度(MLPS,”等保”)与关键信息基础设施(CII)。等保几乎适用于中国境内每一个网络运营者,是合规底线;关键信息基础设施则是在等保之上、针对一旦遭到破坏将危害国家安全、经济或公共利益的系统所施加的更高安全等级。二者均以《网络安全法》(2017 年 6 月 1 日施行)为依据,分别由公安机关(等保)与网信部门(CII 认定)主管。

本文说明等保 2.0 的运作方式、五个保护等级对外资企业的含义、CII 如何认定及其附加义务,以及外资企业落地合规的实务步骤。

等保 2.0:分级保护的底线

等保 2.0 是中国网络安全等级保护制度的现行版本,以国家标准 GB/T 22239-2019《信息安全技术 网络安全等级保护基本要求》 为锚点,该标准于 2019 年 5 月发布、2019 年 12 月 1 日施行,替代 1.0 框架,并将保护范围从传统信息系统扩展至云计算、移动互联、物联网、工业控制系统与大数据。

该制度包含五个步骤:(一)定级——运营者评估系统被破坏后的危害程度并划定等级;(二)备案——将定级结果报送公安机关登记;(三)建设整改——按标准的技术与管理要求补齐短板;(四)等级测评——由具备资质的第三方测评机构验证合规;(五)监督检查——公安机关持续监督。

依《网络安全法》,未落实等保的运营者将面临责令改正、罚款,严重者追究直接责任人员的责任。对外资企业而言,等保合规通常是上线面向用户系统、通过合作方安全审查、以及取得数据出境与许可审批(见相关文章)的前置条件。

五个保护等级

等保按系统被破坏后的危害程度分为五级:

  • 第一级(自主保护级):对公民、法人或其他组织合法权益造成损害,但不损害国家、社会秩序或公共利益,通常自主定级。
  • 第二级(指导保护级):对相关权益造成严重损害,或危害社会秩序、公共利益,但不危害国家安全;须备案,一般定期测评(常见每两年一次)。
  • 第三级(监督保护级):严重危害社会秩序、公共利益,或危害国家安全;有相当规模的商业系统最常见,须备案并每年测评。
  • 第四级(强制保护级):特别严重危害社会秩序、公共利益,或严重危害国家安全——多见于金融、能源、交通核心系统,至少每半年测评一次。
  • 第五级(专控保护级):特别严重危害国家安全——国家专控,民企罕见。

多数外资商业运营落在二级或三级。实务经验:若系统处理大量用户、支付或个人信息,应预期为三级并承担年度测评义务。

关键信息基础设施(CII)

在等保之外,《网络安全法》第三十一条要求国家认定关键信息基础设施——电信、能源、交通、水利、金融、公共服务、电子政务、国防科技工业等重要行业和领域中,一旦遭到破坏、丧失功能或数据泄露,可能严重危害国家安全、国计民生、公共利益 networks and systems。CII 由相关行业主管部门认定、网信办统筹协调。

CII 运营者的义务远超等保:须(一)将在境内收集、产生的重要数据和个人信息存储在境内;(二)任何跨境提供前通过网信办安全评估(见 PIPL 文章);(三)定期开展安全检测与风险评估;(四)仅采购满足 CII 安全要求的产品与服务;(五)遵守专门的 CII 安全保护法规(2021 年国务院《关键信息基础设施安全保护条例》)。处于敏感行业的外资企业——云服务商、支付机构、大型平台、与公用设施关联的业务——应假设自身可能被认定为 CII 并提前规划。

外资企业的义务与执法

对外资企业,关键合规动作包括:确认各系统的中国境内责任主体(通常是境内子公司,而非境外母公司,作为备案主体);就每个面向互联网或承载数据的系统向当地公安机关定级备案;按 GB/T 22239-2019 整改;并委托具备资质的测评机构测评。三级及以上系统须持续年度测评,任何重大变更(新数据类型、新区域、并购)都可能重置定级。

执法真实且日趋活跃。公安机关开展抽查,监管发现会在等保—CII—数据出境体系间共享。不合规可能阻断产品上线、招致罚款,并在涉及个人信息或重要数据时与 PIPL、《数据安全法》相交叠。外资集团应将等保视为由本地问责高管持续负责的常设项目,而非一次性证书。

下一步建议

  • 盘点中国境内系统并为每个系统指定业务责任人;备案主体通常为境内子公司(非总部)。
  • 对照 GB/T 22239-2019 如实定级;面向用户或承载数据的系统应预期二级至三级。
  • 备案与整改并委托具备资质的测评机构,安排周期性测评(三级每年一次)。
  • 评估 CII 暴露面:若经营电信、金融、能源、交通或平台业务,须为境内存储与网信办评估做准备。
  • 建立常态化合规项目(制度、日志、演练)而非一次性审计,并与 PIPL 和数据出境义务统筹。

来源

相关阅读

  • 参见:外资企业的个人信息保护法合规 — CII 与达到阈值处理者须境内存储并通过网信办评估方可出境。
  • 参见:反垄断与经营者集中申报 — 并购可能重置系统定级并触发新备案。
  • 参见:VIE 架构与监管趋势 — 境外上市的数据安全审查。

Cybersecurity Review in China 2026: New Scrutiny for Foreign Security Vendors

  • On 6 August 2026 the Cybersecurity Review Office (under the Cyberspace Administration of China) announced a network-security review of products sold by Palo Alto Networks in China, citing protection of critical information infrastructure.
    2026年8月6日,国家网络安全审查办公室公告,依据《国家安全法》《网络安全法》和《网络安全审查办法》,对派拓公司(Palo Alto Networks)在华销售的产品实施网络安全审查。
  • The review signals heightened scrutiny of foreign cybersecurity vendors whose products plug into China’s critical information infrastructure (CII).
    该审查释放出对深度接入中国关键信息基础设施(CII)的外国网络安全厂商审查趋严的信号。
  • Under the Cybersecurity Review Measures, suppliers of network products and services to CII operators are subject to review; a completed review can lead to procurement restrictions for affected products.
    依据《网络安全审查办法》,向关键信息基础设施运营者提供网络产品和服务的供应商须接受审查;审查结论可导致相关产品采购受限。
  • China’s 2026 measures on cybersecurity and data governance — including the draft rules for large personal-information processors published on 7 August 2026 — continue to raise the compliance bar for foreign technology firms.
    中国2026年的网络安全与数据治理措施——包括2026年8月7日公布的《大型个人信息处理者个人信息保护规定(征求意见稿)》——持续提高外资科技企业的合规门槛。
  • The review comes amid a broader set of reciprocal measures, including export-control adjustments on drones and countermeasures against US entities announced by MOFCOM on 5 August 2026.
    该审查发生在更广泛的对等措施背景下,包括商务部2026年8月5日公布的无人机出口管制调整与针对美国实体的反制措施。
  • For foreign vendors, the practical implication is to design China offerings with review-readiness: data localisation, source-code and architecture transparency, and CII supply-chain documentation.
    对外国厂商而言,实际含义是在设计中国产品时就具备审查就绪性:数据本地化、源代码与架构透明度、以及CII供应链文档。

Cybersecurity Review in China 2026: New Scrutiny for Foreign Security Vendors | 中国2026年网络安全审查:外国安全厂商的新一轮审视

Why this matters now

On 6 August 2026, China’s Cybersecurity Review Office announced a network-security review of products that Palo Alto Networks sells in China. The announcement, published on the Cyberspace Administration of China (CAC) website, invokes the National Security Law, the Cybersecurity Law and the Cybersecurity Review Measures, and states the purpose as protecting critical information infrastructure and guarding against network-security risks. For foreign technology companies — especially cybersecurity vendors, cloud providers and equipment makers whose products sit inside Chinese enterprises’ networks — the review is a concrete reminder that China’s security review regime is an active, operational gate, not a dormant provision.

The review framework at work

The Cybersecurity Review Measures (effective 15 February 2022) establish the legal basis for reviewing network products and services that operators of critical information infrastructure (CII) intend to purchase, or that could affect national security. A review focuses on risks to the security and stable operation of CII; risks of illegal data collection, retention, use, processing and export; risks of products or services being illegally controlled or interrupted; and risks to supply-chain security. Where a review is opened, the outcome can range from conditions attached to procurement to prohibitions on use. The Palo Alto review sits in this framework: a foreign security vendor whose products are deployed in networks that may include CII must now demonstrate that its offering does not create unacceptable risks. For vendors, the review process itself demands cooperation — providing information, materials, technical documents and access for assessment — and non-cooperation can itself lead to a negative conclusion.

The broader 2026 context

The announcement is not isolated. On 7 August 2026, CAC published the draft Provisions on the Protection of Personal Information by Large Personal Information Processors, consolidating two earlier drafts, with a comment deadline of 7 September 2026. The draft extends obligations to processors handling personal information of 10 million or more individuals, deepening the regime that foreign platforms and data-heavy businesses must already navigate. On 5 August 2026, MOFCOM announced tightened export controls on drones and drone-related dual-use items to the United States (Announcement No. 34 of 2026), along with countermeasures against several US entities (Order No. 2 and Order No. 3 of 2026). Taken together, the sequence shows a government that is simultaneously building out its cybersecurity gate, its data-governance rulebook and its reciprocal-trade toolkit — all of which foreign technology firms must treat as a single, connected compliance environment.

Practical steps for foreign technology vendors

  1. Assess CII exposure. Determine whether your products or services are used by operators of critical information infrastructure in China; if so, map which products fall within review scope.
  2. Prepare review documentation. Assemble security designs, source-code access arrangements, data-flow maps, third-party component lists and supply-chain documentation in a review-ready format.
  3. Review data practices. Confirm that products deployed in China comply with personal-information protection and data-export rules, including the emerging large-processor obligations for relevant businesses.
  4. Structure local presence. Where possible, align with data localisation expectations and consider whether a local entity, local storage and local support improve the review posture.
  5. Monitor the political risk layer. Because security reviews can interact with broader reciprocity measures, maintain a watch on MOFCOM export-control and countermeasure announcements that may affect your group.
  6. Build in response discipline. If a review opens, cooperate within statutory timeframes, keep internal records consistent, and engage counsel familiar with the Cybersecurity Review Measures.

Risks and compliance notes

The highest risk for a foreign vendor is complacency: assuming that a sale to a Chinese enterprise is “just a commercial deal” when the product’s deployment may trigger review. Sales into government, finance, energy, transportation and telecom networks — classic CII sectors — deserve the most scrutiny. Also note that the review regime is evolving quickly; the large-processor draft shows that obligations can expand to new categories of actors. Finally, the interconnection with export controls and countermeasures means a vendor’s Chinese business can be affected by measures aimed at its home country; global compliance teams should model those scenarios. None of this argues against the Chinese market — it argues for entering it with the same security-review maturity that the market now expects.


中国2026年网络安全审查:外国安全厂商的新一轮审视

为什么当下重要

2026年8月6日,国家网络安全审查办公室宣布对派拓公司(Palo Alto Networks)在华销售的产品实施网络安全审查。公告发布于中央网信办(国家互联网信息办公室)官网,援引《国家安全法》《网络安全法》和《网络安全审查办法》,目的表述为保障关键信息基础设施安全稳定运行、防范网络安全风险隐患。对外资科技企业——尤其是产品部署在中国企业网络内部的安全厂商、云服务商与设备制造商——这次审查是一个具体提醒:中国的安全审查制度是一道活跃、运转中的闸门,而非沉睡的条款。

运行中的审查框架

《网络安全审查办法》(2022年2月15日施行)为审查关键信息基础设施(CII)运营者拟采购、或可能影响国家安全的网络产品与服务确立了法律依据。审查聚焦以下风险:关键信息基础设施安全稳定运行;非法收集、存储、使用、加工、传输、提供、公开、删除个人信息与重要数据;产品服务被非法控制或干扰;以及供应链安全。一旦启动审查,结论可能从附加采购条件到禁止使用。派拓审查正处在此框架中:产品可能部署于包含CII的网络中的外国安全厂商,现在必须证明其产品不会产生不可接受的风险。对厂商而言,审查过程本身要求配合——提供信息、材料、技术文档并接受评估——不配合本身即可导致不利结论。

2026年更宏观的背景

该公告并非孤立事件。2026年8月7日,国家网信办公布《大型个人信息处理者个人信息保护规定(征求意见稿)》,整合两份早前草案,意见反馈截止2026年9月7日。草案将义务扩展到处理1000万人以上个人信息的处理者,深化了外资平台与数据密集型企业本已必须应对的监管体系。2026年8月5日,商务部公布加强对美国无人机及无人机相关两用物项出口管制(2026年第34号公告),并对多家美国实体采取反制措施(2026年第2号、第3号令)。综合来看,这一系列动作显示政府正同步构建网络安全闸门、数据治理规则与对等贸易工具箱——外资科技企业必须将三者视为一个相互联通的整体合规环境。

外资科技厂商的操作步骤

  1. 评估CII敞口。 确定产品或服务是否被中国关键信息基础设施运营者使用;如是,梳理哪些产品落入审查范围。
  2. 准备审查文档。 以审查就绪格式汇编安全设计、源代码访问安排、数据流图、第三方组件清单与供应链文档。
  3. 复核数据实践。 确认部署于中国的产品符合个人信息保护与数据出境规则,包括相关业务将面对的新兴大型处理者义务。
  4. 构建本地存在。 在可行处与数据本地化预期对齐,评估本地实体、本地存储与本地支持是否改善审查姿态。
  5. 盯住政治风险层。 由于安全审查可能与更广泛的对等措施相互作用,持续关注可能影响集团的商务部出口管制与反制公告。
  6. 建立应对纪律。 一旦启动审查,在法定期限内配合、保持内部记录一致,并聘请熟悉《网络安全审查办法》的法律顾问。

风险与合规提示

外国厂商最大的风险是自满:当产品部署可能触发审查时,却把对中资企业的销售仅当作”一桩商业交易”。政府、金融、能源、交通与电信网络——典型的CII领域——的销售最值得审慎对待。还需注意,审查制度正在快速演进;大型处理者草案表明义务可以扩展到新的主体类别。最后,与出口管制和反制措施的相互关联意味着,厂商的中国业务可能受到针对其母国措施的影响;全球合规团队应建模这些情景。这些都不构成回避中国市场的理由——而是提示要以市场现在所期望的同样的安全审查成熟度进入这个市场。

Sources

  • 中央网信办官网《国家互联网信息办公室关于〈大型个人信息处理者个人信息保护规定(征求意见稿)〉公开征求意见的通知》(2026-08-07,意见反馈截止2026-09-07):https://www.cac.gov.cn/2026-08/07/c_1787851071612596.htm
  • 国家网络安全审查办公室2026-08-06公告:对派拓公司(Palo Alto Networks)在华销售的产品实施网络安全审查(据清华国际金融与经济研究中心2026年中美贸易争端大事记引述原文链接):https://cifer.pbcsf.tsinghua.edu.cn/info/1136/3737.htm
  • 商务部《政策》栏目:商务部公告2026年第34号 公布加强无人机相关两用物项对美国出口管制(2026-08-05);商务部令二〇二六年第2号、第3号 反制决定(2026-08-05):https://www.mofcom.gov.cn/zcfb/index.html
  • 商务部官网《政务公开》:商务部公告2026年第33号 对进口打印复印办公设备发起对外贸易国家安全立案调查(2026-08-05,与网络安全审查同批对等动向):https://www.mofcom.gov.cn/zwgk/index.html

China’s Draft Rule for Large Personal-Information Processors: What Foreign-Invested Platforms and Apps Need to Know

  • China’s cyberspace regulator has released a draft rule for “large personal-information processors,” consolidating two earlier platform-focused drafts into one accountability regime for the biggest data holders.
    国家网信部门发布《大型个人信息处理者个人信息保护规定(征求意见稿)》,将早前两份聚焦平台的规定整合为面向最大规模数据处理者的统一责任制度。
  • A processor handling over 10 million individuals’ personal information, or meeting other significance tests, must self-assess and file a recognition application through its provincial cyberspace authority.
    处理超过1000万人个人信息的处理者,或符合其他重大影响标准的,须自评估并通过省级网信部门申报大型个人信息处理者认定。
  • Core duties include data minimisation, encryption, de-identification, access control, minor-protection safeguards and use of the national online identity authentication public service.
    核心义务涵盖最小必要、加密、去标识化、访问控制、未成年人保护,以及应用国家网络身份认证公共服务。
  • The consultation closes on 7 September 2026, giving foreign-invested platforms and apps a clear compliance horizon to assess exposure.
    征求意见截至2026年9月7日,为外资平台与App提供了评估合规敞口的明确窗口。
  • The draft sits atop the Personal Information Protection Law and the Network Data Security Management Regulation, signalling converging oversight of super-platforms.
    征求意见稿置于《个人信息保护法》与《网络数据安全管理条例》之上,预示着对超级平台的趋同监管。
  • Practical step: map data-volume and business-footprint triggers now, and pre-position a recognition filing before the threshold is crossed.
    实操建议:立即对照数据量与业务足迹触发条件制图,并在越过门槛前预先准备认定申报。

China’s Draft Rule for Large Personal-Information Processors: What Foreign-Invested Platforms and Apps Need to Know | 大型个人信息处理者规定征求意见:外资平台与App须知

On 7 August 2026, the Cyberspace Administration of China (CAC) published the *Draft Provisions on the Personal Information Protection of Large Personal-Information Processors (for public comment)*. The draft consolidates two earlier consultation documents — the rules on establishing personal-information protection supervisory committees for large online platforms and the rules on large online platforms’ personal-information protection — into a single, broader regime that reaches any “large personal-information processor,” not only classic internet platforms.

For foreign-invested businesses operating consumer apps, e-commerce storefronts, ride-hailing or delivery services, social products, or any service that scales to millions of Chinese users, the draft is a material compliance event. This article explains who is captured, what changes, and the practical steps to take during the consultation window.

Who is a “large processor”

The draft applies a multi-factor test rather than a single mechanical threshold. Under Article 2, a “large personal-information processor” is a processor whose personal-information protection and supervision are subject to the provisions. Recognition is triggered when any of the following conditions is met:

  1. Scale: the processor handles the personal information of more than 10 million natural persons.
  2. Importance of service: it provides important online services involving personal-information processing, or its business scope covers multiple lines of business that involve personal-information processing.
  3. Impact: its personal-information processing activities have an important influence on national security, economic operation, social stability, public health and safety, or similar interests.

Article 3 obliges processors that handle more than 10 million individuals’ data, and those that self-assess into the second or third condition, to file a recognition application through the provincial cyberspace authority to the national cyberspace authority. Provincial authorities must complete a completeness check within 15 working days of receiving the materials; if complete, they forward the materials and a preliminary opinion to the national authority. Authorities may also urge a processor that meets the conditions but has not filed to do so.

Key compliance duties

The draft imposes obligations that go well beyond the baseline PIPL duties:

  • Data minimisation. Collection must be limited to what is necessary to provide the product or service; over-collection is prohibited.
  • Technical and organisational safeguards. Processors should adopt encryption, de-identification, access control and anonymisation, may apply the national data-label/identification technology, and are encouraged to obtain personal-information protection certification.
  • Minor protection. Processing personal information of minors under 14 requires the consent of a parent or guardian, and convenient channels must be provided for giving and withdrawing that consent.
  • National identity service. Processors are encouraged to use the national online identity authentication public service.
  • Monitoring and response. Processors must build internal management systems and operating procedures covering processing activities, rights-response and security obligations, strengthen monitoring and early warning of personal-information security, and detect and dispose of security incidents in a timely manner.
  • International engagement. The draft encourages large processors to participate in international standards and rule-making on personal-information protection and to promote mutual recognition of rules and standards with other countries and regions.

Consultation timeline and process

The CAC notice sets the comment deadline at 7 September 2026. Feedback may be submitted by email to shujuju@cac.gov.cn or by post to the Cyberspace Administration of China’s Network Data Administration Bureau in Beijing, with the envelope marked “大型个人信息处理者个人信息保护规定征求意见.” Because the draft is still at the consultation stage, the final obligations, recognition form and timetable may shift — but the direction is clear: accountability for the largest data holders is being formalised.

Practical steps for foreign-invested platforms

  1. Size the footprint. Quantify the number of natural persons whose personal information you process in China; if you are near or above 10 million, assume you are in scope.
  2. Assess the qualitative triggers. Even below 10 million, multi-line consumer businesses with systemic reach should self-assess the “importance” and “impact” prongs.
  3. Pre-position the filing. Prepare the recognition application materials (data-scale evidence, processing-activity description, safeguard design) so a filing can be lodged promptly if triggered.
  4. Harden the baseline. Encrypt and de-identify at rest and in transit, restrict access on least-privilege, and stand up a minor-consent workflow.
  5. Engage the consultation. Where group policy allows, submit comments through the recognised industry association channel before 7 September 2026.

大型个人信息处理者规定征求意见:外资平台与App须知

2026年8月7日,国家互联网信息办公室发布《大型个人信息处理者个人信息保护规定(征求意见稿)》。该稿将早前两份文件——关于大型网络平台设立个人信息保护监督委员会的规定、以及关于大型网络平台个人信息保护的规定——整合为一项覆盖任何”大型个人信息处理者”、而非仅限传统互联网平台的统一制度。

对于运营消费类App、电商店面、网约车或配送服务、社交产品,或任何在中国拥有数百万用户规模的外商投资企业而言,该征求意见稿是一项重要的合规事件。本文说明谁被纳入、有何变化,以及在征求意见窗口期应采取的实操步骤。

何为”大型个人信息处理者”

该稿采用的是多因素测试,而非单一机械门槛。根据第二条,大型个人信息处理者的个人信息保护及其监督管理适用本规定。符合下列条件之一的,应认定属于大型个人信息处理者:

  1. 规模: 处理1000万人以上自然人个人信息。
  2. 服务重要性: 提供涉及个人信息处理的重要网络服务,或者经营范围涵盖多种业务且涉及个人信息处理。
  3. 影响: 个人信息处理活动对国家安全、经济运行、社会稳定、公共健康和安全等具有重要影响。

第三条规定,处理1000万人以上自然人个人信息,以及自评估认为符合第二条第二款第二、三项条件的处理者,应当通过所在地省级网信部门向国家网信部门申报大型个人信息处理者认定并按要求提交材料。省级网信部门应自收到申报材料之日起15个工作日内完成完备性查验;材料齐全的,将材料与初步认定意见报送国家网信部门。省级以上网信、电信主管部门、公安机关等部门认为符合认定条件但未主动申报的,应督促其申报。

核心合规义务

该稿施加了超出《个人信息保护法》基线的义务:

  • 最小必要。 收集个人信息应限于提供产品或者服务所必需,不得过度收集。
  • 技术与组织保障。 采取加密、去标识化、访问控制、匿名化等措施;可应用国家数据标签标识技术;鼓励通过个人信息保护认证。
  • 未成年人保护。 处理不满十四周岁未成年人个人信息应取得监护人同意,并提供便捷的同意与撤回渠道。
  • 国家身份服务。 鼓励应用国家网络身份认证公共服务。
  • 监测与响应。 围绕处理活动、权利响应、安全义务建立健全内部管理制度与操作规程,加强监测预警,及时发现处置安全事件。
  • 国际参与。 鼓励参与个人信息保护国际标准和规则制定,推动与其他国家、地区规则标准协调互认。

征求意见时间与程序

网信办通知设定意见反馈截止时间为2026年9月7日。可通过电子邮件发送至 shujuju@cac.gov.cn,或邮寄至北京市海淀区阜成路15号国家互联网信息办公室网络数据管理局(信封注明”大型个人信息处理者个人信息保护规定征求意见”)。该稿尚处征求意见阶段,最终义务、认定表单与时间表可能调整,但方向明确:最大数据持有者的责任正在制度化。

外资平台实操步骤

  1. 测算足迹。 量化在华处理的自然人个人信息数量;若接近或高于1000万,应假定自身落入范围。
  2. 评估定性触发条件。 即便低于1000万,具有系统性触达的多线消费业务也应自评估”重要性”与”影响”两项。
  3. 预先准备申报。 备妥认定申报材料(数据规模证据、处理活动说明、保障措施设计),以便在触发时及时提交。
  4. 夯实基线。 对静态与传输中数据加密、去标识化,按最小权限限制访问,并建立未成年人同意工作流。
  5. 参与征求意见。 在集团政策允许范围内,于2026年9月7日前通过认可的行业协会议事渠道提交意见。

Practitioner Quick Reference

  • One regime, broader reach: the draft folds two platform rules into a single “large processor” regime covering 10M+ individuals or systemic-impact handlers.
    一项制度、更广覆盖:该稿将两份平台规则整合为覆盖1000万以上或具系统影响处理者的”大型处理者”制度。
  • File through the province: recognition applications go via the provincial cyberspace authority to the national CAC; completeness check is 15 working days.
    经省级申报:认定申请由省级网信部门报送国家网信办;完备性查验为15个工作日。
  • Deadline 7 September 2026: the consultation window is the time to assess and, if eligible, comment.
    截止2026年9月7日:征求意见窗口是评估并(如符合)提交意见的时机。
  • Baseline must rise: minimisation, encryption, de-identification, access control and minor-consent are explicit duties.
    基线须抬高:最小必要、加密、去标识化、访问控制与未成年人同意均为明示义务。

Common Pitfalls

  • Assuming “platform” only means social media: the test reaches any processor at 10M individuals or with systemic impact, including e-commerce and on-demand apps.
    误以为”平台”仅指社交媒体:测试覆盖任意达1000万或具系统影响的处理者,含电商与按需服务App。
  • Waiting for the final text: the recognition filing and safeguard build-out take months; starting during consultation is cheaper than scrambling after promulgation.
    等待终稿:认定申报与保障体系建设需数月;在征求意见期启动比颁布后仓促应对更划算。
  • Treating PIPL compliance as sufficient: the draft adds duties (minimisation, national identity service, minor workflows) beyond the PIPL baseline.
    以为符合个保法即足够:该稿新增了超出个保法基线的义务(最小必要、国家身份服务、未成年人流程)。

Sources

Cybersecurity Review and Critical Information Infrastructure Procurement Compliance for Foreign-Invested Enterprises

  • The Cybersecurity Review regime applies whenever a Critical Information Infrastructure (CII) operator procures network products or services that may affect national security — and foreign-invested enterprises supplying those products are squarely within scope.
    当关键信息基础设施(CII)运营者采购可能影响国家安全的网络产品与服务时,网络安全审查制度即被触发;向这些运营者供应产品的外商投资企业同样被直接纳入监管视野。
  • The legal backbone is the Cybersecurity Law (2017), the Regulation on CII Protection (State Council Decree No. 745, 2021) and the Cybersecurity Review Measures (CAC Decree No. 8, 2022), now reinforced by active enforcement against foreign-made security products in 2026.
    制度主干为《网络安全法》(2017)、《关键信息基础设施安全保护条例》(国务院令745号,2021)与《网络安全审查办法》(国家网信办等13部门令第8号,2022),并在2026年针对境外安全产品的执法中持续强化。
  • CII operators must “prioritise the procurement of secure and trustworthy” products and submit to security review before signing; using unreviewed or failed products triggers fines of 1–10 times the procurement value.
    CII运营者须”优先采购安全可信的网络产品”,并在签约前申报审查;使用未申报或审查未通过的产品,将面临采购金额一倍以上十倍以下的罚款。
  • A supplier need not be a CII operator itself — providing products or services to a CII operator can independently trigger a review, as the August 2026 review of a US network-security vendor’s China-sold products illustrates.
    供应商本身不必是CII运营者——仅为CII运营者提供产品即可能单独触发审查,2026年8月对某美国网络安全厂商在华销售产品的审查即为例证。
  • Foreign-invested enterprises should embed review-trigger analysis into procurement playbooks, vendor due diligence and contract clauses, and retain evidence of supplier commitments not to extract user data or interrupt supply.
    外商投资企业应将”审查触发”分析嵌入采购流程、供应商尽调与合同条款,并留存供应商”不非法获取用户数据、不中断供应”的承诺证据。
  • The review is risk-based and time-bound (generally 30 + 15 working days, with a special procedure extending to about 90), and explicitly states it is not intended to restrict or discriminate against foreign products and services.
    审查以风险为本且有明确时限(一般30+15个工作日,特别审查最长约90个工作日),并明确规定其目的不是限制或歧视国外产品与服务。

Cybersecurity Review and Critical Information Infrastructure Procurement Compliance for Foreign-Invested Enterprises | 网络安全审查与关键信息基础设施采购合规:外商投资企业的实务指南

Why this matters now

For a foreign-invested enterprise (FIE) selling network equipment, cloud services, security software or industrial control components into the China market, the question is no longer whether cybersecurity rules apply, but *when a procurement crosses the line into a national-security review*. China’s cybersecurity framework has matured from a set of principles into an operational, enforcement-driven regime. The clearest recent signal came on 6 August 2026, when the Cybersecurity Review Office announced a review of products sold in China by a US network-security vendor — the second time, after the 2023 review of a US memory-chip maker, that the measure was applied to specific foreign-made products rather than to entities. For FIEs on either side of a CII procurement — as the operator, the supplier, or both — the compliance calculus has sharpened.

This article explains the statutory triggers, the three pillars of the legal framework, what the review actually assesses, the 2026 enforcement signal, and the concrete steps an FIE should take. It does not address the separate (though related) regimes of data-export security assessment, standard contracts and certification, which are covered elsewhere on this site.

The statutory trigger: who and when

The trigger is defined in Article 2 of the Cybersecurity Review Measures: a CII operator procuring network products and services, or a platform operator conducting data-processing activities, that *may affect or may affect* national security must undergo cybersecurity review. The wording is deliberately forward-looking — the operator must pre-assess the risk before signing and file for review where the risk exists.

Two points matter for foreign suppliers in particular. First, the obligation to file sits with the CII operator (the buyer), not the supplier. Second, and often overlooked, a supplier can be pulled into a review even when it is not itself a CII operator: providing products or services *to* a CII operator is exactly the scenario the 2026 review targeted. The Measures (Article 21) also state that where foreign-investment security review or data-security review have separate rules, those apply alongside.

CII itself is defined broadly in the Regulation on CII Protection: public communications and information services, energy, transport, water conservancy, finance, public services, e-government, defence science and industry, and other networks/systems whose disruption or data breach would seriously endanger national security, the public interest or social stability. Telecom, finance, energy, transport, healthcare and government IT are the obvious sectors — precisely the sectors where foreign technology is most often deployed.

The three pillars of the legal framework

Pillar one — the Cybersecurity Law (adopted 2016, effective 2017). Article 35 imposes the core duty: a CII operator procuring network products and services that may affect national security shall pass the security review organised by the cyberspace authority and relevant departments. Article 37 separately restricts CII operators from storing personal information and important data overseas or providing it abroad without assessment.

Pillar two — the Regulation on CII Protection (State Council Decree No. 745, effective 1 September 2021). This fills in the duty-holder detail: operators must build a dedicated security-management organ, conduct at least one security testing and risk assessment per year, and — critically for procurement — “prioritise the procurement of secure and trustworthy network products and services” and put any potentially security-affecting procurement through the national cybersecurity review. Article 19 makes this a positive legal obligation, not mere guidance.

Pillar three — the Cybersecurity Review Measures (CAC Decree No. 8, effective 15 February 2022). Thirteen departments jointly issued the Measures, which specify the review objects (core network devices, important communications products, high-performance computers and servers, large-capacity storage, large databases and application software, network-security devices, cloud-computing services, and other products/services materially affecting CII, network and data security). They set the 45-working-day baseline (30 + 15, with a special procedure of about 90 working days) and the penalty ladder.

What the review assesses

Article 10 of the Measures lists the national-security risk factors the review weighs:

  • the risk that products/services, once in use, lead to illegal control of, interference with or destruction of CII;
  • the risk that supply interruption harms the business continuity of CII;
  • the security, openness, transparency and source diversity of the product/service, and the risk of supply interruption for political, diplomatic or trade reasons;
  • the provider’s compliance with Chinese laws, administrative regulations and departmental rules;
  • the risk of core data, important data or large volumes of personal information being stolen, leaked, destroyed, illegally used or illegally exported;
  • for listings abroad, the risk that CII, core data, important data or large volumes of personal information fall under foreign-government influence or control; and
  • other factors endangering CII, network and data security.

For a foreign supplier, the most relevant factors are supply-chain continuity, transparency of the product’s architecture, and — increasingly — whether the product could become a vector for data extraction or remote control. The 2026 enforcement action against a foreign security vendor centred precisely on the paradox that a security product can itself become a security risk if compromised or externally controlled.

The 2026 enforcement signal

On 6 August 2026 the Cybersecurity Review Office announced a review, under the Cybersecurity Law and the Measures, of products sold in China by a US network-security company, citing the need to safeguard CII and prevent cybersecurity risks. Chinese regulators had, earlier in 2026, directed domestic institutions to stop using certain US and Israeli security products and to complete migration to domestic alternatives within the first half of the year. The pattern is consistent: reviews now target *products* (not just entities), focus on CII supply-chain security, and operate alongside a broader “secure and trustworthy” procurement preference in sensitive sectors.

The practical takeaway for FIEs is not panic but preparation. The Measures explicitly state the review is “not intended to restrict or discriminate against foreign products and services,” and China continues to welcome foreign products into its market. The differentiator is demonstrable supply-chain resilience, source transparency and contractual commitments.

Practical compliance steps for FIEs

  1. Map your exposure. Identify whether your customers include CII operators or platform operators handling large-scale personal information. If so, your products are in the review’s line of sight even when you are not the filer.
  2. Build review-trigger analysis into the sales cycle. Equip local teams to recognise when a customer’s procurement will require a filing, and pre-empt delays by surfacing the issue early.
  3. Harden the supplier commitments. The Measures let CII operators require providers to promise not to illegally obtain user data, not to illegally control or manipulate user equipment, and not to interrupt supply without just cause. Put these in writing and be able to evidence them.
  4. Localise where it reduces risk. Source diversity, in-country support, transparent architecture and local data handling are exactly the factors the review rewards. A wholly import-dependent supply chain is the weakest posture.
  5. Keep contracts review-aware. Where a contract is signed before review, include a condition that it takes effect only after the product passes review, to avoid the “use unreviewed product” penalty of 1–10 times procurement value.
  6. Coordinate with the FIE’s data-export posture. Cybersecurity review overlaps with data-export security assessment and PIPL compliance; treat them as one programme rather than three silos.

Related reading

  • Critical information infrastructure recognition and the annual self-assessment obligation.
  • Data-export pathways: security assessment, standard contract and certification compared.
  • PIPL compliance for foreign-invested enterprises operating in China.

网络安全审查与关键信息基础设施采购合规:外商投资企业的实务指南

为何重要

对于向中国市销售网络设备、云服务、安全软件或工业控制部件的外商投资企业(FIE)而言,问题已不再是”网络安全规则是否适用”,而是”一项采购何时越过红线、进入国家安全审查”。中国的网络安全框架已从原则性规定,发展为可操作、强执法的制度体系。最清晰的近期信号出现在2026年8月6日——网络安全审查办公室宣布对某美国网络安全厂商在华销售产品启动审查,这是继2023年对某美国存储芯片企业之后,第二次将审查直接指向特定境外制造的产品而非实体。无论外商投资企业处于CII采购的哪一端——作为运营者、供应商或兼而有之——合规逻辑都已显著收紧。

本文阐释法定触发条件、法律框架的三大支柱、审查究竟评估什么、2026年的执法信号,以及外商投资企业应采取的具体步骤。数据出境安全评估、标准合同与认证等相对独立(但相关)的制度,不在本文讨论范围,本站点另有专文。

法定触发:谁、何时

触发条件规定于《网络安全审查办法》第二条:关键信息基础设施运营者采购网络产品与服务,或网络平台运营者开展可能影响国家安全的数据处理活动,应当进行网络安全审查。措辞具有前瞻性——运营者须在签约前预判风险,并在风险存在时申报审查。

有两点对外国供应商尤为重要。其一,申报义务落在CII运营者(买方)身上,而非供应商。其二(也最常被忽视):供应商即便自身不是CII运营者,也可能被卷入审查——仅为CII运营者提供产品,正是2026年审查所针对的情形。《办法》第二十一条还规定,外商投资安全审查、数据安全审查另有规定的,应一并适用。

CII的定义见于《关键信息基础设施安全保护条例》,范围宽泛:公共通信和信息服务、能源、交通、水利、金融、公共服务、电子政务、国防科技工业,以及其他一旦遭到破坏、丧失功能或数据泄露可能严重危害国家安全、公共利益或社会稳定的网络设施与信息系统。电信、金融、能源、交通、医疗与政府IT是最典型的行业——也正是外国技术最常部署的领域。

法律框架的三大支柱

支柱一——《网络安全法》(2016通过,2017施行)。 第三十五条规定核心义务:CII运营者采购可能影响国家安全的网络产品与服务,应当通过国家网信部门会同有关部门组织的国家安全审查。第三十七条另限制CII运营者未经评估不得将个人信息与重要数据境外存储或提供境外。

支柱二——《关键信息基础设施安全保护条例》(国务院令745号,2021年9月1日施行)。 细化义务主体:运营者须设立专门安全管理机构、每年至少开展一次网络安全检测与风险评估,且——对采购尤为关键——”优先采购安全可信的网络产品”,并将任何可能影响安全的采购纳入国家网络安全审查。第十九条将其确立为积极的法定义务,而非软性指引。

支柱三——《网络安全审查办法》(国家网信办等13部门令第8号,2022年2月15日施行)。 该办法明确审查对象(核心网络设备、重要通信产品、高性能计算机与服务器、大容量存储设备、大型数据库与应用软件、网络安全设备、云计算服务,以及对CII、网络与数据安全有重要影响的其他产品服务),设定45个工作日的基准时限(30+15,特别审查约90个工作日),并规定罚则梯度。

审查评估什么

《办法》第十条列举审查权衡的国家安全风险因素:

  • 产品服务投入使用后导致CII被非法控制、干扰或破坏的风险;
  • 供应中断损害CII业务连续性的风险;
  • 产品服务的安全性、开放性、透明性、来源多样性,以及因政治、外交、贸易因素导致供应中断的风险;
  • 提供者遵守中国法律、行政法规与部门规章的情况;
  • 核心数据、重要数据或大量个人信息被窃取、泄露、毁损、非法利用或非法出境的风险;
  • 赴国外上市情形下,CII、核心数据、重要数据或大量个人信息受外国政府影响或控制的风险;
  • 其他危害CII、网络与数据安全的因素。

对外国供应商而言,最相关的因素依次是供应链连续性、产品架构的透明度,以及——日益重要——产品是否会成为数据提取或远程控制的载体。2026年对某外国安全厂商的执法,恰恰聚焦于”安全产品本身可能因其被控制或存在后门而转化为安全风险”这一悖论。

2026年的执法信号

2026年8月6日,网络安全审查办公室依据《网络安全法》与《办法》,宣布对某美国网络安全公司在华销售产品启动审查,理由为保障CII安全、防范网络安全风险。在此之前的2026年内,中国监管部门已要求境内机构停止使用该厂商等部分美国、以色列安全产品,并在上半年内完成向国产替代的迁移。脉络一致:审查如今指向*产品*(而非仅实体),聚焦CII供应链安全,并与敏感行业”安全可信”采购偏好并行推进。

对外商投资企业的实务启示不是恐慌,而是准备。《办法》明确审查”不是要限制或歧视国外产品和服务”,中国仍欢迎国外产品进入市场。差异点在于可证明的供应链韧性、来源透明度与合同承诺。

外商投资企业的实务步骤

  1. 梳理暴露面。 识别客户是否包含CII运营者或处理大规模个人信息的平台运营者。若是,即便你不是申报方,产品也已进入审查视野。
  2. 将审查触发分析嵌入销售周期。 让本地团队能够识别客户采购何时需要申报,并及早暴露问题以避免延误。
  3. 夯实供应商承诺。 《办法》允许CII运营者要求提供者承诺不非法获取用户数据、不非法控制或操纵用户设备、无正当理由不中断供应。将这些承诺书面化并备可佐证材料。
  4. 以本地化降低风险。 来源多样性、本地支持、透明架构与本地数据处理,正是审查所鼓励的因素。完全依赖进口供应链是最弱姿态。
  5. 让合同”审查友好”。 若合同在审查前签署,应附条件约定”经审查通过后方可生效”,以规避”使用未审查产品”面临的采购金额一倍以上十倍以下罚款。
  6. 与数据出境合规协同。 网络安全审查与数据出境安全评估、个保法合规相互重叠,应作为一个整体项目而非三座孤岛来管理。

延伸阅读

  • 关键信息基础设施认定与年度自评估义务。
  • 数据出境路径比较:安全评估、标准合同与认证。
  • 外商投资企业在华的个保法合规。

Sources

  • 国家互联网信息办公室等十三部门:《网络安全审查办法》(令 第8号,2022年2月15日施行)— https://www.cac.gov.cn/2022-01/04/c_1642894602182845.htm
  • 《关键信息基础设施安全保护条例》(国务院令 第745号,2021年9月1日施行)— https://www.cac.gov.cn/2021-08/17/c_1630785976988160.htm
  • 《中华人民共和国网络安全法》(主席令 第五十三号,2017年6月1日施行)— http://www.npc.gov.cn/npc/c2/c12435/c12488/201905/t20190521_270249.html

Data Classification, Grading and Critical Information Infrastructure Obligations for Foreign-Invested Enterprises

  • China mandates a data classification and grading (数据分类分级) baseline; the national standard GB/T 43697-2024 took effect on 1 October 2024.
    中国确立了数据分类分级基线制度,国家标准GB/T 43697-2024于2024年10月1日施行。
  • The Network Data Security Management Regulations (State Council Order No. 790, effective 1 January 2025) require handlers to classify and grade network data and identify important data.
    《网络数据安全管理条例》(国务院令790号,2025年1月1日施行)要求处理者对网络数据分类分级并识别重要数据。
  • Data is graded into core, important and general; important data triggers catalogue reporting and stricter export control.
    数据分为核心、重要、一般三级;重要数据触发目录报送与更严格的出境管控。
  • A critical information infrastructure operator (CIIO) owes extra duties: a security assessment, a person in charge, and annual risk assessment.
    关键信息基础设施运营者承担额外义务:安全评估、安全负责人与年度风险评估。
  • The 2026 financial-information sector guide shows the model: three primary classes, four grades, and a 67-item third-level list.
    2026年金融信息服务数据分类分级指南给出范本:3个一级类、4级、67项三级分类。
  • Cross-border data flows are gated by grade: important data needs a security assessment; unidentified data generally does not.
    数据出境按级别设闸:重要数据须安全评估,未被认定为重要数据的通常无需申报。

Data Classification, Grading and Critical Information Infrastructure Obligations for Foreign-Invested Enterprises | 外资企业的数据分类分级与关键信息基础设施义务

The Baseline: Classify and Grade

China’s data-governance stack is built on one foundational duty: data classification and grading (数据分类分级). The Data Security Law (2021) created the system; the Cybersecurity Law, the PIPL and the Network Data Security Management Regulations (State Council Order No. 790, effective 1 January 2025) give it teeth. The national standard GB/T 43697-2024, Data Security Technology — Rules for Data Classification and Grading, released on 21 March 2024 and effective 1 October 2024, is the general rulebook that industry regulators and handlers follow.

How Grading Works

Under GB/T 43697-2024, data is graded by its importance to economic and social development and the harm that leakage, tampering, destruction or unlawful use would cause. The framework has three tiers:

  • Core data: covers areas or groups with high coverage or precision/scale/depth, whose unlawful use could directly affect political security—e.g. data on national security key areas, the national economic lifeline, major public welfare or major public interest.
  • Important data: specific fields, groups or regions, or data of a certain precision and scale, whose alteration or leakage could directly endanger national security, economic operation, social stability, public health or safety.
  • General data: everything else, further sub-divided by handlers as needed (often 2–4 internal levels).

Key principles: scientific and practical, clear boundaries, grade up where in doubt (就高从严), point-and-area combined, and dynamic updating.

What “Important Data” Actually Means for an FIE

Important-data identification is the practical flashpoint. The 2026 CAC guidance on the financial-information sector illustrates the live model: data is classified by business attribute into three primary classes (business, user, enterprise), then sub-divided into nine secondary and 67 tertiary classes; it is graded into core, important, sensitive-general and routine-general. Handlers follow a defined flow—data inventory, classification, grading, a graded catalogue, important-data directory reporting, and dynamic updating.

For an ordinary FIE, the safe reading is: you are expected to run a data inventory and apply the grading rules yourself, and to report important data to the competent department once identified. Under the Provisions on Promoting and Regulating Cross-Border Data Flows, data not notified or published as important by a department or region does not require a security assessment—but the action obligation to classify and identify is not waived.

Critical Information Infrastructure (CII) Duties

If a handler is designated a critical information infrastructure operator (CIIO), the obligations rise sharply:

  • a security assessment before commencing operations and on material change;
  • a person in charge of network security and a dedicated management body;
  • an annual risk-assessment report on network data processing;
  • stricter rules on data location and on cross-border transfer of personal information and important data, which generally requires a CAC security assessment.

CIIO status is determined by the relevant industry regulator; an FIE in telecom, finance, energy, healthcare, transport or cloud/services with national-scale reach should assume it may be in scope.

The Network Data Security Management Regulations in Practice

State Council Order No. 790 operationalises the above: it sets the classify-and-grade duty, the important-data catalogue and reporting obligation, the CIIO security assessment and person-in-charge requirements, and the dedicated duties of large platform providers (e.g. an easy-to-find option to turn off personalised recommendation, and an annual personal-information protection social-responsibility report for the largest platforms). It also optimises the cross-border mechanism: a handler may export personal information where it meets the conditions, and data not notified as important need not be assessed as such.

Why This Matters for Cross-Border Flows

Classification and grading is the gate to every export route. The chain is: inventory → classify → grade → identify important data → if important, security assessment before export; if general and within the standard-contract or certification bands, the lighter routes apply; if a CIIO, assessment is the default. An FIE that has not graded its data cannot know which route it needs, and mis-grading important data as general is a direct violation.

Practical Steps for a Foreign-Invested Enterprise

  1. Inventory the data you hold in China by business scene and entity.
  2. Apply GB/T 43697-2024 to classify (by industry/business attribute) and grade (core/important/general).
  3. Identify and report important data to the competent department; keep the catalogue dynamic.
  4. Check CIIO status with your industry regulator; if in scope, build the assessment, person-in-charge and annual-report files.
  5. Map each data stream to its export route before any transfer; do not move data you have not graded.

Related Reading

See our cross-border data transfer rules (No. 35), the field-level data export negative list (No. 57) and the amended Cybersecurity Law in force from 2026 (No. 178).


外资企业的数据分类分级与关键信息基础设施义务

基线:分类与分级

中国数据治理体系的基石是一项基础义务——数据分类分级。2021年《数据安全法》建立了制度;《网络安全法》《个人信息保护法》与《网络数据安全管理条例》(国务院令790号,2025年1月1日施行)赋予其执行力。国家标准GB/T 43697-2024《数据安全技术 数据分类分级规则》于2024年3月21日发布、2024年10月1日施行,是行业主管部门与数据处理者遵循的通用规则。

分级如何运作

根据GB/T 43697-2024,数据依其在经济社会发展中的重要程度,以及遭篡改、破坏、泄露或非法使用所造成的危害进行分级,框架分三级:

  • 核心数据:对领域、群体、区域具有较高覆盖度或达到较高精度、较大规模、一定深度,非法使用可能直接影响政治安全——如国家安全重点领域、国民经济命脉、重大民生、重大公共利益相关数据。
  • 重要数据:特定领域、特定群体、特定区域或达到一定精度规模,遭篡改或泄露可能直接危害国家安全、经济运行、社会稳定、公共健康与安全的数据。
  • 一般数据:其余数据,由处理者按需再细分(通常2—4级)。

基本原则:科学实用、边界清晰、就高从严、点面结合、动态更新。

“重要数据”对外资企业的实际含义

重要数据识别是实务焦点。网信办2026年金融信息服务领域指南给出了鲜活范本:数据按业务属性分为业务、用户、企业3个一级类,再细分9个二级、67个三级类;按核心、重要、敏感一般、常规一般分为4级。处理者按既定流程作业——数据梳理、分类、分级、形成分级清单、报送重要数据目录、动态更新。

对普通外资企业,稳妥理解是:应自行开展数据盘点并套用分级规则,并在识别到重要数据后向主管部门报送。依据《促进和规范数据跨境流动规定》,未被相关部门、地区告知或公开发布为重要数据的,无需作为重要数据申报安全评估——但分类分级与识别的”作为”义务并不因此免除。

关键信息基础设施(CII)义务

若被认定为关键信息基础设施运营者(CIIO),义务显著加重:

  • 运营前及重大变更前开展安全评估;
  • 设置网络安全负责人与专门管理机构;
  • 每年提交网络数据处理风险评估报告;
  • 数据与系统的存放及个人信息、重要数据出境受到更严约束,通常须通过网信部门安全评估。

CIIO身份由相关行业主管部门认定;在电信、金融、能源、医疗、交通或云/服务领域具有全国规模的外资企业,应假定自己可能落入范围。

《网络数据安全管理条例》的落地

国务院令790号将上述要求落地:确立分类分级义务、重要数据目录与报送义务、CIIO安全评估与负责人要求,以及大型平台企业的专有责任(如提供易于查找的个性化推荐关闭选项、最大平台每年发布个人信息保护社会责任报告)。同时优化了跨境机制:处理者符合条件即可出境个人信息;未被通知为重要数据的,无需作为重要数据申报评估。

为何这关系到跨境流动

分类分级是每条出境路径的闸门。链条为:盘点→分类→分级→识别重要数据→若为重要数据,出境前安全评估;若为一般数据且落在标准合同或认证区间,走较轻路径;若为CIIO,评估为默认。未分级的外资企业无从判断该走哪条路;把重要数据误判为一般数据,属直接违规。

外资企业的实操步骤

  1. 盘点数据:按业务场景与主体梳理在华持有的数据。
  2. 套用GB/T 43697-2024:按行业/业务属性分类,按核心/重要/一般分级。
  3. 识别并报送重要数据给主管部门;目录保持动态更新。
  4. 核实CIIO身份:与行业主管部门确认;若落入范围,建立评估、负责人与年度报告档案。
  5. 每条数据流映射至出境路径后再传输;未分级的数据不要动。

延伸阅读

可参阅本站跨境数据传输规则(编号35)、字段级数据出境负面清单(编号57)与2026年施行的修订《网络安全法》(编号178)。

Sources

China’s amended Cybersecurity Law in force from 1 January 2026: AI clauses, tenfold penalties and wider extraterritorial reach

  • The Decision amending the Cybersecurity Law was adopted on 28 October 2025 and takes effect on 1 January 2026, with 14 amendments and a reissued text.
    修改《网络安全法》的决定于2025年10月28日通过、2026年1月1日起施行,共14项修改并重新公布。
  • A new Article 20 builds the first basic-law “development plus oversight” framework for artificial intelligence in China.
    新增第20条,在中国基础性法律层面首次为人工智能构建”发展保障+安全监管”双重框架。
  • Maximum fines jump from the million-yuan range to RMB 10 million, and dual liability on units and responsible individuals is now the norm.
    罚款上限由百万级跃升至1000万元级,单位与直接责任人的双罚制成常态。
  • Critical information infrastructure operators face procurement-linked penalties of one to ten times the purchase amount.
    关键信息基础设施运营者面临”采购金额1—10倍”的罚则。
  • Extraterritorial application is widened: overseas actors harming China’s cybersecurity may be pursued and sanctioned, including asset freezes.
    域外适用扩大,境外危害中国网络安全的机构、组织、个人可被追究法律责任并面临包括冻结财产在内的制裁。
  • Personal information handling is explicitly tied to the Civil Code and PIPL, tightening the legal basis for processing.
    个人信息处理被明确要求遵守《民法典》《个人信息保护法》,做实法律衔接。
  • Foreign-invested enterprises should re-screen CII status, verify product certifications, and re-align cross-border data and remote-access arrangements.
    外商投资企业应重新梳理关键信息基础设施认定、核验产品认证,并对齐跨境数据与远程访问安排。

China’s amended Cybersecurity Law in force from 1 January 2026: AI clauses, tenfold penalties and wider extraterritorial reach | 2026年1月1日施行的新修改《网络安全法》:人工智能专条、阶梯式重罚与域外适用扩围

Overview: Why This Matters to Foreign-Invested Enterprises

The *Decision of the Standing Committee of the National People’s Congress on Amending the Cybersecurity Law* was adopted at the 18th session of the 14th NPC Standing Committee on 28 October 2025, promulgated by Presidential Order No. 61, and takes effect on 1 January 2026, with 14 amendments and a reissued text.

The Cybersecurity Law, the Data Security Law and the Personal Information Protection Law form the backbone of China’s network and data security legal system. The amendment raises the fine ceiling by an order of magnitude, introduces the first basic-law AI framework, tightens the Civil Code/PIPL link, and expands extraterritorial reach.

The Legislative Facts

The amending Decision was passed at the 18th meeting of the Standing Committee of the 14th NPC on 28 October 2025, promulgated by Presidential Order No. 61, and takes effect on 1 January 2026; the original law had been in force since 1 June 2017, and the amendment renumbers the articles and reissues the full text. Per the Constitution and Law Committee report of 27 October 2025, the Standing Committee deliberated on 25 October, the Committee studied the articles clause by clause that evening, and — after consulting relevant departments — fixed the effective date as 1 January 2026, adopting the suggestions to add AI-promotion content and to widen extraterritorial application.

A New Guiding Principle and the AI Article

The first amendment inserts a new Article 3: “Cybersecurity work adheres to the leadership of the Communist Party of China, implements the holistic view of national security, coordinates development and security, and advances the building of a cyber power.”

The headline addition is new Article 20 — the AI clause. It provides that the state supports foundational AI theoretical research and R&D of key technologies such as algorithms, promotes training-data and computing-power infrastructure, improves AI ethics norms, strengthens risk monitoring, assessment and security supervision, and promotes sound AI application and development; it also supports innovative cybersecurity management and the use of new technologies such as AI to raise protection levels.

Personal Information Linkage

The former Article 40 becomes new Article 42 and adds a second paragraph: a network operator processing personal information must comply with this Law and with laws and administrative regulations such as the Civil Code and the Personal Information Protection Law.

Stepped Penalties: The Order-of-Magnitude Jump

The former Article 59 becomes new Article 61 with a graded, stepped schedule; the maximum fine rises to RMB 10 million.

SeverityWhoUnit fine (RMB)Individual fine (RMB)
General duty failure (Art. 23/27), persistsNetwork operator50k–500k10k–100k
CII duty failure (Art. 35/36/38/40), persistsCII operator100k–1m10k–100k
Serious: large leak / CII partial lossEither500k–2m50k–200k
Especially serious: CII main-function lossEither2m–10m200k–1m

The design is graduated: a first-tier failure draws a warning and a modest fine; only persistence or actual harm escalates. The upper tiers — large-scale data leakage or a CII losing partial function, and a CII losing its main function — carry the RMB 10 million ceiling and individual liability up to RMB 1 million.

Product Certification and the Closing of Applications

New Article 63 targets the supply side: selling or providing network key equipment or dedicated cybersecurity products that have not passed security certification or detection, or that are unqualified, draws an order to stop sales or provision, a warning and confiscation of unlawful gains; where gains are below RMB 100,000 an additional RMB 20,000–100,000 fine applies, and where gains exceed RMB 100,000 an additional one-to-five-times fine applies; in serious cases, business may be suspended, rectified, or licences revoked.

The former Article 61 becomes new Article 64, and disposal measures add “closing the application” alongside “closing the website”.

Critical Information Infrastructure: Procurement and Offshore Storage

The former Article 65 becomes new Article 67: a CII operator using products or services not security-reviewed, or that failed review, in violation of Article 37, is ordered to rectify, stop use, eliminate effects on national security, and fined one to ten times the procurement amount, with responsible individuals fined RMB 10,000–100,000.

New Article 65 (former Article 62) covers unauthorised cybersecurity certification, detection, risk assessment, or publicly releasing vulnerabilities, viruses, attack or intrusion information: a warning and up to RMB 100,000; persistent or serious cases draw RMB 100,000–1,000,000 plus possible suspension, closure or licence revocation, with individuals fined up to RMB 100,000.

New Article 69 (merging former Articles 68 and 69(1)) covers failure to stop transmission of unlawful information, take elimination measures, keep records or report: a warning and up to RMB 500,000; persistent or serious cases draw RMB 500,000–2,000,000 plus possible suspension, closure or revocation, with individuals fined RMB 50,000–200,000; especially serious impact raises the fine to RMB 2,000,000–10,000,000 and individuals to RMB 200,000–1,000,000; electronic-information and app-download providers failing Article 50(2) duties are punished under the same paragraphs.

New Article 71 (merging former Articles 64, 66 and 70) routes three categories to other laws: (1) publishing or transmitting information prohibited under Article 13(2) and other laws; (2) infringing personal-information rights under Articles 24(3) and 43–45; and (3) a CII operator storing personal information and important data offshore, or providing such data offshore, in violation of Article 39; stealing or illegally obtaining, selling or providing personal information not yet constituting a crime is punished by public security authorities.

Proportionate Penalties and Extraterritorial Reach

New Article 73 introduces a general leniency clause: where a violation also meets circumstances for lighter, mitigated or exempt punishment under the Administrative Penalty Law, those provisions apply.

The final amendment changes the former Article 75 to new Article 77 and widens extraterritorial application. Overseas institutions, organisations or individuals endangering China’s cybersecurity are now “pursued for legal liability according to law”; where serious consequences are caused, the public security department and relevant State Council departments “may decide to take freezing of assets or other necessary sanctions.” Adding “pursued for legal liability” and broadening the trigger materially expands the law’s reach beyond China’s borders.

How Officials Framed the Revision

The CAC Network Rule-of-Law Bureau described the amended law as enriching guiding principles, adding AI security and development content, aligning personal-information protection with the Civil Code and PIPL, improving the liability system, and widening extraterritorial application. He Bo, director of the CAICT Internet Law Research Center, noted the stepped penalties categorise general, persistent/serious and especially serious cases, applying punishment and education so enforcement “has both strength and warmth,” and called the 1 January 2026 effective date a marker that “China’s cybersecurity rule-of-law has entered a new stage”; the NPC Standing Committee Economic Law Office stated the amendment implements the thought on building a cyber power and that the three laws together constitute China’s network and data security legal system.

What Foreign-Invested Enterprises Should Do Next

  1. Re-screen CII status. Determine whether your China entity, or parts of it, are or may be designated critical information infrastructure, since the heaviest tiers and the procurement multiple attach to that status.
  2. Verify product certifications. Audit network key equipment and dedicated cybersecurity products against the certification/detection catalogues; obtain and retain proof before purchase and deployment.
  3. Re-align the legal basis for personal-information processing. Update privacy notices, processing records and contracts to cite Article 42(2) and the Civil Code/PIPL linkage.
  4. Build an approval path for outward vulnerability and security-information release. Article 65 rules require internal sign-off before any external disclosure.
  5. Review group-level cross-border data and remote access. The Article 71 offshore route and widened Article 77 reach bear directly on group clouds, cross-border systems and remote operations by overseas parent or affiliate staff.
  6. Re-rate the compliance budget and individual exposure. With dual liability normalised and a RMB 10 million ceiling, assign board-level ownership and director-officer risk awareness.

Related Reading

Two companion articles on this site cover adjacent ground without overlap: the Multi-Level Protection Scheme (MLPS) and foreign-invested enterprises, and the Regulations on Network Data Security Management; the cross-references are pointers.


2026年1月1日施行的新修改《网络安全法》:人工智能专条、阶梯式重罚与域外适用扩围

概览:本次修法对外商投资企业的含义

《全国人民代表大会常务委员会关于修改〈中华人民共和国网络安全法〉的决定》于2025年10月28日第十四届全国人大常委会第十八次会议通过,经国家主席令第六十一号公布,自2026年1月1日起施行,共14项修改并重新公布全文。

《网络安全法》与《数据安全法》《个人信息保护法》共同构成我国网络与数据安全法律体系的支柱;本次修法将罚款上限提高一个数量级,首次在基础性法律层面为人工智能设立框架,做实与《民法典》《个人信息保护法》的衔接,并扩大域外适用。

立法事实

修改决定于2025年10月28日第十四届全国人大常委会第十八次会议通过,由中华人民共和国主席令第六十一号公布,自2026年1月1日起施行。原法自2017年6月1日施行;本次修改相应调整条文顺序并重新公布全文。根据2025年10月27日宪法和法律委员会的报告,常委会于10月25日分组审议,宪法和法律委员会当晚逐条研究,并与有关部门研究后将施行时间确定为2026年1月1日;委员会采纳了”增加促进人工智能应用、运用人工智能提升网络安全保护水平”的意见,也采纳了”扩大域外适用情形”的建议。

新增指导原则与人工智能专条

第一项修改新增第3条:”网络安全工作坚持中国共产党的领导,贯彻总体国家安全观,统筹发展和安全,推进网络强国建设。”

实质性新增是第20条(人工智能专条)。其规定,国家支持人工智能基础理论研究和算法等关键技术研发,推进训练数据资源、算力等基础设施建设,完善人工智能伦理规范,加强风险监测评估和安全监管,促进人工智能应用和健康发展;同时规定,国家支持创新网络安全管理方式,运用人工智能等新技术,提升网络安全保护水平。这是中国基础性法律层面首次为人工智能构建”发展保障+安全监管”双重框架。

个人信息保护的衔接

修改将原第40条改为第42条并增加一款:网络运营者处理个人信息,应当遵守本法和《民法典》《个人信息保护法》等法律、行政法规的规定。这做实了与《民法典》《个人信息保护法》的衔接。

阶梯式罚则:量级的跃升

原第59条改为第61条,引入分级、阶梯式处罚安排,罚款上限由百万级跃升至1000万元级。

情节层级适用主体单位罚款(元)责任人罚款(元)
一般性义务未履行且拒不改正(第23、27条)网络运营者5万–50万1万–10万
关基义务未履行且拒不改正(第35、36、38、40条)关基运营者10万–100万1万–10万
严重后果:大量数据泄露、关基丧失局部功能二者均适用50万–200万5万–20万
特别严重后果:关基丧失主要功能二者均适用200万–1000万20万–100万

设计刻意呈阶梯状:第一层级仅招致警告与较低罚款;唯有拒不改正或实际造成危害才逐级抬升。但顶层——大规模数据泄露或关基丧失局部功能,以及关基丧失主要功能的特别严重情形——触及1000万元上限,责任人个人责任可达100万元。

产品认证检测与应用关停

新增第63条指向供给端:销售或提供未经安全认证、安全检测,或者不合格、不符合要求的网络关键设备和网络安全专用产品,将被责令停止销售或提供、警告并没收违法所得;无违法所得或不足10万元的,并处2万–10万元;违法所得10万元以上的,并处违法所得1–5倍;情节严重的,可责令暂停业务、停业整顿、吊销许可证或营业执照。

原第61条改为第64条,处置措施在”关闭网站”之外新增”关闭网站或者应用程序”。

关键信息基础设施:采购与境外存储

原第65条改为第67条:关基运营者违反第37条,使用未经安全审查或安全审查未通过的网络产品或者服务的,被责令限期改正、停止使用、消除对国家安全的影响,并处采购金额1–10倍罚款,责任人处1万–10万元。

新增第65条(原第62条)规制未经许可开展网络安全认证、检测、风险评估,或向社会发布系统漏洞、计算机病毒、网络攻击、网络侵入等网络安全信息:责令改正、警告,可处1万–10万元;拒不改正或情节严重的,处10万–100万元,并可责令暂停业务、停业整顿、关闭网站或应用程序、吊销许可证或营业执照,责任人处1万–10万元。

新增第69条(合并原第68条与第69条第1项)规制未停止传输违法信息、未采取消除处置措施、未保存记录或未报告:责令改正、警告、通报,可处5万–50万元;拒不改正或情节严重的,处50万–200万元,并可责令暂停业务、停业整顿、关闭网站或应用程序、吊销许可证或营业执照,责任人处5万–20万元;造成特别严重影响、特别严重后果的,处200万–1000万元,责任人处20万–100万元。电子信息发送服务提供者、应用软件下载服务提供者不履行第50条第2款安全管理义务的,依照前两款处罚。

新增第71条(合并原第64、66、70条)将三类行为交由其他法律、行政法规处理处罚:(一)发布或传输第13条第2款和其他法律禁止发布或传输的信息;(二)违反第24条第3款、第43至45条规定侵害个人信息权益;(三)违反第39条规定,关基运营者在境外存储个人信息和重要数据,或向境外提供个人信息和重要数据。窃取或非法获取、非法出售或非法提供个人信息尚不构成犯罪的,由公安机关依法处罚。

过罚相当与域外适用扩围

新增第73条引入一般从轻条款:违反本法规定但具有《行政处罚法》规定的从轻、减轻或不予处罚情形的,依照其规定。

最后一项修改将原第75条改为第77条,扩大域外适用。其现规定:境外的机构、组织、个人从事危害中华人民共和国网络安全活动的,”依法追究法律责任”;造成严重后果的,国务院公安部门和有关部门并可以决定对该机构、组织、个人采取冻结财产或者其他必要的制裁措施。原条文仅规定冻结财产等制裁,新增”依法追究法律责任”且触发情形扩大,使该法的域外辐射显著拓宽。

官方对本次修法的定调

中央网信办网络法治局指出,修改后的网络安全法”充实网络安全工作指导原则,增加促进人工智能安全与发展的内容,在个人信息保护方面做好与《民法典》和《个人信息保护法》的衔接,进一步完善网络安全法律责任制度,扩大域外适用情形”。中国信通院互联网法律研究中心主任何波在网信办官网刊文表示,新法设置阶梯式处罚,按一般性违法、拒不改正或情节严重、造成特别严重影响/后果分类施策,宽严相济,使执法”既有力度又有温度”;并称2026年1月1日施行”标志着我国网络安全法治建设进入新阶段”。全国人大常委会法工委经济法室表示,修改重点之一是贯彻网络强国重要思想、充实指导原则;网络安全法与数据安全法、个人信息保护法共同构成我国网络安全和数据安全法律体系。

外商投资企业下一步应做什么

  1. 重新梳理关基认定。 厘清中国实体或其部分是否被认定或可能被认定为关键信息基础设施,因为最重罚则层级与采购倍数均附着于该身份。
  2. 核验产品认证。 对照认证/检测目录审计网络关键设备与网络安全专用产品,在采购与部署前取得并留存证明。
  3. 对齐个人信息处理的法律依据。 更新隐私政策、处理记录与合同,引据第42条第2款及《民法典》《个人信息保护法》衔接。
  4. 建立漏洞与网络安全信息对外发布的审批路径。 第65条风险评估与信息发布规则要求任何对外披露前完成内部签批。
  5. 复核集团层面的跨境数据与远程访问。 第71条境外存储/提供路径与拓宽的第77条域外适用,直接关涉集团云、跨境系统与境外母公司或关联方人员的远程运维。
  6. 重估合规预算与个人风险。 双罚制常态化、上限1000万元,须配置董事会层级的责任归属与高管风险意识。

延伸阅读

本站另有两篇相邻主题文章,互不重叠:网络安全等级保护(MLPS)与外资企业、《网络数据安全管理条例》。新修改的《网络安全法》作为基础性法律居于二者之上;此处仅作交叉引用,不作重写。

Sources