一号台 · 登记册附件Desk #1 · Register annex

AI 产品出海与入华规则清单Rules Checklist for AI Products Going Global and Entering China

本清单是《AI 产品出海登记册》(EXP)的附件,也是「出海准入自检」的题源底稿。登记对象是 AI 产品跨越市场边界时会撞上的规则本身:准入、数据、内容、分发与资本各条线上的法规与规章,逐项只认官方原文。This checklist is an annex to the AI Global Expansion Register (EXP), and the source base for the expansion readiness self-check. It registers the rules themselves that an AI product runs into when it crosses a market border: the laws and regulations on the access, data, content, distribution and capital lines, each verified against official texts only.

编制口径:每项规则只认官方原文(监管机构官网、官方公报、法规数据库原页),逐项打开核对。出台不等于生效,生效不等于执法已动。执法案例只作条目下的「执法注脚」,不混进规则本体。找不到官方原文或官方原文支撑不了的,一律标待核并写明缺口,不删项、不拿媒体转述充一手。本清单只记规则事实与官方口径,不作法律判断,不构成合规意见。How it is compiled: every rule is taken from official texts only (regulator websites, official gazettes, official legal database pages), opened and checked one by one. Promulgation is not entry into force, and entry into force is not enforcement. Enforcement cases appear only as enforcement footnotes under an entry, never inside the rule itself. Anything that no official text could be found for, or that the official text cannot carry, is marked pending verification with the gap stated; nothing is deleted, and no media report is passed off as a primary source. This checklist records rule facts and official positions only. It makes no legal judgment and is not compliance advice.

归属Desk一号台(企业 AI 架构情报台)· EXP 登记册附件Desk 1 (Enterprise AI Architecture) · Annex to the EXP register版本Versionv0.1(2026-11-01 首编)v0.1 (first edition, November 1, 2026)核查截至Checked as of2026-11-01署名By深翼 DeepWingDeepWing

状态四格怎么读How to read the four status cells

已出台:规则文本已经通过并公布,记通过与公布日期。已生效:规则已经生效或开始适用,分阶段者逐段记适用日,修订改过的日期与原始日期分开记。执法已动:执法机制已建或已有执法动作,分开记机制与个案。待核:官方原文没核到或支撑不了的部分,写明缺口,不作定论。Promulgated: the text has been adopted and published, with its adoption and publication dates. In force: the rule is in force or has begun to apply; phased application is recorded stage by stage, and dates changed by amendment are kept apart from the original dates. Enforcement: an enforcement mechanism exists or enforcement action has been taken, with mechanism and individual cases recorded separately. Pending: what official texts could not be found for, or cannot carry, with the gap stated and no conclusion drawn.

每项另记对应 EXP 卡点(一至二,取登记册八类卡点原名)、对 AI 产品的实际关口与官方来源链接。Each entry also records its EXP blocker categories (one or two, using the register's eight category names as written), the practical gate for AI products, and links to the official sources.

一、出海向:目的市场规则Part I · Outbound: destination-market rules

R-01欧盟《人工智能法案》EU Artificial Intelligence Act已生效 · 分阶段适用In force · phased application R-02欧盟《通用数据保护条例》(GDPR)EU General Data Protection Regulation (GDPR)已生效 · 执法已动In force · enforcement active R-03欧盟《数字服务法》(DSA)EU Digital Services Act (DSA)已生效 · 执法已动In force · enforcement active R-04科罗拉多 AI 法Colorado AI Act已出台 · 过渡状态待核Promulgated · transitional status pending R-05得克萨斯州 TRAIGATexas TRAIGA已生效(2026-01-01)· 执法机制已建In force (Jan 1, 2026) · enforcement mechanism in place R-06加州 CCPA(经 CPRA 修订)California CCPA (as amended by the CPRA)已生效 · 执法已动In force · enforcement active R-07加州 AB 2013 训练数据透明California AB 2013, Training Data Transparency已出台 · 义务自 2026-01-01 起算Promulgated · duties run from Jan 1, 2026 R-08加州 SB 942 AI 透明法California SB 942, AI Transparency Act已施行(2026-08-02)· SB 1000 修订已即时生效Operative (Aug 2, 2026) · SB 1000 amendments in immediate effect R-09FTC 法第五条FTC Act Section 5现行有效 · 执法已动且口径有回调In force · enforcement active, posture partly rolled back R-10韩国个人信息保护法(PIPA)Korea's Personal Information Protection Act (PIPA)已生效In force R-11韩国《人工智能基本法》Korea's Framework Act on Artificial Intelligence已生效(2026-01-22)· 罚款设宽限期In force (Jan 22, 2026) · fine grace period R-12日本个人信息保护法(APPI)Japan's Act on the Protection of Personal Information (APPI)已生效 · 滚动修订中In force · under rolling amendment R-13巴西通用个人数据保护法(LGPD)Brazil's General Personal Data Protection Law (LGPD)已生效(分段)In force (in stages) R-14印度数字个人数据保护法(DPDP Act)India's Digital Personal Data Protection Act (DPDP Act)分三段指定生效 · 核心义务在第三段Commencement in three tranches · core duties in the third

二、中国出境门Part II · China's outbound gate

R-15数据出境安全评估办法Measures for Data Export Security Assessment已生效 · 阈值以 R-17 现行口径为准In force · thresholds read under R-17 R-16个人信息出境标准合同办法Measures for the Standard Contract for Personal Information Export已生效 · 阈值以 R-17 现行口径为准In force · thresholds read under R-17 R-17促进和规范数据跨境流动规定Provisions on Promoting and Regulating Cross-Border Data Flows已生效 · 现行分档口径以本规定为准In force · the current tiering follows this instrument R-18出口管制法Export Control Law已生效 · 与 AI 的具体列管对应待核In force · specific AI listing correspondence pending

三、入华向Part III · Inbound to China

R-19生成式人工智能服务管理暂行办法Interim Measures for the Management of Generative AI Services已生效 · 备案持续公布In force · filings published continuously R-20互联网信息服务算法推荐管理规定Provisions on the Management of Algorithmic Recommendations in Internet Information Services已生效 · 算法备案公示中In force · algorithm filings public R-21互联网信息服务深度合成管理规定Provisions on the Management of Deep Synthesis in Internet Information Services已生效 · 备案批次持续公布In force · filing batches published continuously R-22人工智能生成合成内容标识办法Measures for Labeling AI-Generated and Synthesized Content已生效(2025-09-01)· 执法通报已出In force (Sep 1, 2025) · enforcement notice issued R-23网络安全法(含 2025 年修正)Cybersecurity Law (including the 2025 amendment)已生效 · 2025 年修正决定 2026-01-01 施行In force · the 2025 amendment decision takes effect Jan 1, 2026 R-24数据安全法Data Security Law已生效In force R-25个人信息保护法Personal Information Protection Law已生效 · 有域外适用条款In force · with extraterritorial application

四、待核项清单(缺口与原因)Part IV · Pending-verification annex (items and the reasons for their gaps)

R-04科罗拉多 AI 法Colorado AI Act已出台 · 过渡状态待核Promulgated · transitional status pending R-05得克萨斯州 TRAIGATexas TRAIGA已生效(2026-01-01)· 执法机制已建In force (Jan 1, 2026) · enforcement mechanism in place R-06加州 CCPA(经 CPRA 修订)California CCPA (as amended by the CPRA)已生效 · 执法已动In force · enforcement active R-07加州 AB 2013 训练数据透明California AB 2013, Training Data Transparency已出台 · 义务自 2026-01-01 起算Promulgated · duties run from Jan 1, 2026 R-09FTC 法第五条FTC Act Section 5现行有效 · 执法已动且口径有回调In force · enforcement active, posture partly rolled back R-10韩国个人信息保护法(PIPA)Korea's Personal Information Protection Act (PIPA)已生效In force R-11韩国《人工智能基本法》Korea's Framework Act on Artificial Intelligence已生效(2026-01-22)· 罚款设宽限期In force (Jan 22, 2026) · fine grace period R-12日本个人信息保护法(APPI)Japan's Act on the Protection of Personal Information (APPI)已生效 · 滚动修订中In force · under rolling amendment R-13巴西通用个人数据保护法(LGPD)Brazil's General Personal Data Protection Law (LGPD)已生效(分段)In force (in stages) R-14印度数字个人数据保护法(DPDP Act)India's Digital Personal Data Protection Act (DPDP Act)分三段指定生效 · 核心义务在第三段Commencement in three tranches · core duties in the third R-16个人信息出境标准合同办法Measures for the Standard Contract for Personal Information Export已生效 · 阈值以 R-17 现行口径为准In force · thresholds read under R-17 R-17促进和规范数据跨境流动规定Provisions on Promoting and Regulating Cross-Border Data Flows已生效 · 现行分档口径以本规定为准In force · the current tiering follows this instrument R-18出口管制法Export Control Law已生效 · 与 AI 的具体列管对应待核In force · specific AI listing correspondence pending

一、出海向:目的市场规则Part I · Outbound: destination-market rules

R-01已生效 · 分阶段适用In force · phased application

欧盟《人工智能法案》EU Artificial Intelligence Act

规则名Official nameRegulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act),中文通称欧盟《人工智能法案》Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)

法域Jurisdiction欧盟(条例,全部成员国直接适用)European Union (a regulation, directly applicable in all member states)

方向Direction出海Outbound

对应卡点EXP blocker监管审批Regulatory approval

已出台Promulgated

2024 年 6 月 13 日通过,2024 年 7 月 12 日刊于欧盟官方公报。Adopted on June 13, 2024, and published in the Official Journal of the European Union on July 12, 2024.

已生效In force

2024 年 8 月 1 日生效(原文 Article 113,公布后第 20 日)。原始文本的分阶段适用日逐个列明:主体适用日 2026 年 8 月 2 日;第一章、第二章(含 Article 5 禁止性实践)自 2025 年 2 月 2 日适用;第三章第四节、第五章(通用目的 AI 模型义务)、第七章、第十二章与 Article 78 自 2025 年 8 月 2 日适用(Article 101 除外);Article 6(1) 及对应义务自 2027 年 8 月 2 日适用(附件 I 高风险过渡)。Entered into force on August 1, 2024 (Article 113 of the text, the twentieth day after publication). The original text's phased application dates, listed one by one: the main application date is August 2, 2026; Chapters I and II (including the Article 5 prohibited practices) apply from February 2, 2025; Chapter III Section 4, Chapter V (general-purpose AI model obligations), Chapter VII, Chapter XII and Article 78 apply from August 2, 2025 (Article 101 excepted); Article 6(1) and its corresponding obligations apply from August 2, 2027 (the Annex I high-risk transition).

执法已动Enforcement

部分已动。禁止性实践与通用目的 AI 义务已过适用日;罚则在条文内落地,Article 99 定禁止性实践最高罚 3500 万欧元或全球年营业额 7% 取高者,通用目的 AI 罚则见 Article 101。Partly in motion. The prohibited practices and the general-purpose AI obligations are past their application dates; the penalty provisions are in the text: Article 99 sets the top fine for prohibited practices at EUR 35 million or 7% of total worldwide annual turnover, whichever is higher, and the general-purpose AI penalties are at Article 101.

待核Pending

无日期待核;引用时必须注明所据为合并版(2026 年 7 月 27 日版)。No date is pending; citations must state that they rest on the consolidated version (the version of July 27, 2026).

重要修订(与原始日期分开记)Key amendment (recorded apart from the original dates)

EUR-Lex 合并版(版本日 2026 年 7 月 27 日)显示,Regulation (EU) 2026/1744(Digital Omnibus on AI)已修订 Article 113:第三章第 1 至 3 节改为 2027 年 12 月 2 日适用(Article 6(2) 与附件 III 高风险)、2028 年 8 月 2 日适用(Article 6(1) 与附件 I 高风险);Article 5 新增部分点自 2026 年 12 月 2 日适用;Articles 102 至 110 自 2026 年 7 月 27 日适用。对外引用高风险合规截止日须用合并版新日期,不得再把原始日期写成现行截止日。The EUR-Lex consolidated version (version date July 27, 2026) shows that Regulation (EU) 2026/1744 (the Digital Omnibus on AI) has amended Article 113: Chapter III Sections 1 to 3 now apply from December 2, 2027 (Article 6(2) and Annex III high-risk) and from August 2, 2028 (Article 6(1) and Annex I high-risk); the new points added to Article 5 apply from December 2, 2026; Articles 102 to 110 apply from July 27, 2026. Any outward citation of a high-risk compliance deadline must use the consolidated version's new dates; the original dates must no longer be written as current deadlines.

实际关口Practical gate在欧盟投放市场或投入使用的 AI 系统与通用目的 AI 模型,不论提供者设立在欧盟内还是第三国均适用(原文 Article 2(1)),按禁止性实践、高风险要求、透明度与通用模型分级承担义务。AI systems and general-purpose AI models placed on the EU market or put into service in the EU are covered whether the provider is established inside the EU or in a third country (Article 2(1) of the text), with obligations tiered across prohibited practices, high-risk requirements, transparency and general-purpose models.

官方来源Official sources

核查日期 2026-11-01Checked November 1, 2026

R-02已生效 · 执法已动In force · enforcement active

欧盟《通用数据保护条例》(GDPR)EU General Data Protection Regulation (GDPR)

规则名Official nameRegulation (EU) 2016/679 (General Data Protection Regulation),中文通称《通用数据保护条例》Regulation (EU) 2016/679 (General Data Protection Regulation)

法域Jurisdiction欧盟(条例,全部成员国直接适用)European Union (a regulation, directly applicable in all member states)

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

2016 年 4 月 27 日通过,2016 年 5 月 4 日刊于官方公报 OJ L 119。Adopted on April 27, 2016, and published in the Official Journal, OJ L 119, on May 4, 2016.

已生效In force

Article 99(1) 定公布后第 20 日生效,Article 99(2) 定自 2018 年 5 月 25 日起适用,无再分阶段。附带程序规则 Regulation (EU) 2025/2518 的跨境执法程序自 2027 年 4 月 2 日适用,不影响主体适用日。Article 99(1) sets entry into force on the twentieth day after publication, and Article 99(2) sets application from May 25, 2018, with no further phasing. The cross-border enforcement procedure under the accompanying procedural regulation, Regulation (EU) 2025/2518, applies from April 2, 2027, and does not change the main application date.

执法已动Enforcement

有。各成员国设有完全独立的监管机构负责监督与执法;Article 83 定行政罚款最高 2000 万欧元或企业全球年营业额 4% 取高者。Yes. Each member state has a fully independent supervisory authority responsible for monitoring and enforcement; Article 83 sets administrative fines at up to EUR 20 million or 4% of the undertaking's total worldwide annual turnover, whichever is higher.

待核Pending

无。None.

实际关口Practical gate设立在欧盟外的公司向欧盟内个人提供商品或服务、或监控其行为时同样适用;AI 产品处理欧盟个人数据须满足合法基础、数据主体权利与设计和默认保护等要求。Companies established outside the EU are equally covered when they offer goods or services to individuals in the EU or monitor their behaviour. An AI product processing EU personal data must meet the lawful-basis, data-subject-rights and data-protection-by-design-and-by-default requirements.

官方来源Official sources

核查日期 2026-11-01Checked November 1, 2026

执法注脚(与规则本体分开记)Enforcement footnote (recorded apart from the rule itself)

意大利 Garante 对 DeepSeek 的正式决定(Provvedimento del 30 gennaio 2025,决定登记号第 33 号,doc. web n. 10098477),依 GDPR 第 58(2)(f) 条命令最终限制在意大利境内的个人数据处理、即时生效。该决定已在 EXP-003 条目内以监管原文核实。The Italian Garante's formal decision on DeepSeek (Provvedimento del 30 gennaio 2025, decision register no. 33, doc. web n. 10098477), ordering, under GDPR Article 58(2)(f), a definitive limitation on the processing of personal data in Italy, effective immediately. The decision has been verified against the regulator's original text inside entry EXP-003.

R-03已生效 · 执法已动In force · enforcement active

欧盟《数字服务法》(DSA)EU Digital Services Act (DSA)

规则名Official nameRegulation (EU) 2022/2065 on a Single Market For Digital Services (Digital Services Act),中文通称《数字服务法》Regulation (EU) 2022/2065 on a Single Market For Digital Services (Digital Services Act)

法域Jurisdiction欧盟(条例,全部成员国直接适用)European Union (a regulation, directly applicable in all member states)

方向Direction出海Outbound

对应卡点EXP blocker分发渠道Distribution channels

已出台Promulgated

2022 年 10 月 19 日通过,2022 年 10 月 27 日刊于官方公报 OJ L 277。Adopted on October 19, 2022, and published in the Official Journal, OJ L 277, on October 27, 2022.

已生效In force

Article 93(1) 定公布后第 20 日生效,即 2022 年 11 月 16 日。分阶段适用:主体自 2024 年 2 月 17 日适用;Article 24(2)(3)(6)、Article 33(3) 至 (6)、Article 37(7)、Article 40(13)、Article 43 与第四章第 4 至 6 节(超大平台与搜索引擎相关报告、审计、数据共享与监督规则)自 2022 年 11 月 16 日先行适用。Article 93(1) sets entry into force on the twentieth day after publication, that is, November 16, 2022. Application is phased: the main body applies from February 17, 2024; Article 24(2)(3)(6), Article 33(3) to (6), Article 37(7), Article 40(13), Article 43 and Chapter IV Sections 4 to 6 (the reporting, audit, data-sharing and supervision rules for very large platforms and search engines) applied earlier, from November 16, 2022.

执法已动Enforcement

有。欧盟委员会官网载,委员会于 2026 年 5 月 28 日依本法对 Temu 罚款 2 亿欧元,理由为未勤勉识别与评估其平台非法产品的系统性风险,该案程序于 2024 年 10 月 31 日正式开启。Yes. The European Commission's website records that on May 28, 2026 the Commission fined Temu EUR 200 million under this Act, for failing to diligently identify and assess the systemic risks of illegal products on its platform; the proceedings in that case were formally opened on October 31, 2024.

待核Pending

无。None.

实际关口Practical gate适用于向欧盟内接收者提供的中介服务,不论提供者设立地;在线平台类 AI 产品须履行非法内容通知与处理、透明度与商家可追溯义务,达超大平台门槛者另须做系统性风险评估与缓解。It applies to intermediary services offered to recipients in the EU, wherever the provider is established. Platform-type AI products must run illegal-content notice-and-action, transparency and trader-traceability duties; those reaching the very large platform threshold must additionally assess and mitigate systemic risks.

官方来源Official sources

核查日期 2026-11-01Checked November 1, 2026

R-04已出台 · 过渡状态待核Promulgated · transitional status pending

科罗拉多 AI 法Colorado AI Act

规则名Official nameSB24-205 Consumer Protections for Artificial Intelligence,通称 Colorado AI Act(科罗拉多人工智能法)SB24-205 Consumer Protections for Artificial Intelligence, commonly the Colorado AI Act

法域Jurisdiction美国科罗拉多州State of Colorado, United States

方向Direction出海Outbound

对应卡点EXP blocker监管审批Regulatory approval

已出台Promulgated

州长于 2024 年 5 月 17 日批准,州议会官网法案页记为 APPROVED by Governor May 17, 2024,并记 Became Law。Approved by the Governor on May 17, 2024; the General Assembly's official bill page records "APPROVED by Governor May 17, 2024" and records that it became law.

已生效In force

法案页记 EFFECTIVE May 17, 2024,但实体义务原本自 2026 年 2 月 1 日起适用;SB 25B-004 于 2025 年 8 月 28 日经州长批准,其官方摘要明示将 SB 24-205 各项要求的生效日延至 2026 年 6 月 30 日;其后 SB 26-189 已出台,官方摘要明示其废止并重订(repeals and reenacts)原条文,改为自动决策技术(ADMT)新规,开发者与部署者义务自 2027 年 1 月 1 日起。引用时不得再单写「2026 年 2 月 1 日生效」。The bill page records "EFFECTIVE May 17, 2024", but the substantive duties were originally to apply from February 1, 2026. SB 25B-004, approved by the Governor on August 28, 2025, states in its official summary that it delays the effective date of SB 24-205's requirements to June 30, 2026. After that, SB 26-189 was enacted; its official summary states that it repeals and reenacts the original provisions as a new automated decision-making technology (ADMT) regime, with developer and deployer duties from January 1, 2027. Citations must no longer write "effective February 1, 2026" on its own.

执法已动Enforcement

原法授予州总检察长专属执法权,违反构成科罗拉多消费者保护法下的欺骗性交易行为;本次打开的官方页面未见已发布的执法个案。The original Act gives the state Attorney General exclusive enforcement authority, and a violation is a deceptive trade practice under the Colorado Consumer Protection Act. The official pages opened in this round showed no published enforcement case.

待核Pending

2026 年 6 月 30 日至 2027 年 1 月 1 日之间原条文的实际可执行状态,及涉原法的联邦诉讼对执行的影响,未打开官方过渡条款与法院文件核实,标待核。The actual enforceability of the original provisions between June 30, 2026 and January 1, 2027, and the effect of federal litigation over the original Act on its enforcement, were not verified against the official transitional provisions or court documents, and are marked pending.

实际关口Practical gate原文口径下,用于重大决定(就业、住房、信贷、保险、医疗、教育等)的高风险 AI,开发者与部署者须尽合理注意防止算法歧视,并承担影响评估、风险管理、消费者通知与人工复核申诉等义务;2027 年起按 SB 26-189 的 ADMT 通知与说明口径执行。Under the original text, developers and deployers of high-risk AI used in consequential decisions (employment, housing, credit, insurance, health care, education and the like) must use reasonable care to prevent algorithmic discrimination, and carry impact-assessment, risk-management, consumer-notice and human-review appeal duties. From 2027 the SB 26-189 ADMT notice and explanation regime governs instead.

R-05已生效(2026-01-01)· 执法机制已建In force (Jan 1, 2026) · enforcement mechanism in place

得克萨斯州 TRAIGATexas TRAIGA

规则名Official nameTexas Responsible Artificial Intelligence Governance Act(得克萨斯负责任人工智能治理法,TRAIGA)。编号口径须纠正:成法载体是 HB 149,其 enrolled 原文第 1 条自称本名;HB 1709 是 2024 年 12 月提出的引入版,官方进程止于 2025 年 3 月 14 日提交委员会,未成法,两者不得混写。Texas Responsible Artificial Intelligence Governance Act (TRAIGA). A numbering correction is required: the enacted vehicle is HB 149, whose enrolled text names the Act in its Section 1. HB 1709 is the introduced version filed in December 2024; the official record stops at its referral to committee on March 14, 2025, it was never enacted, and the two must not be conflated.

法域Jurisdiction美国得克萨斯州State of Texas, United States

方向Direction出海Outbound

对应卡点EXP blocker政府及机构使用限制 / 监管审批Government and institutional use restrictions / Regulatory approval

已出台Promulgated

HB 149 于 2025 年 6 月 22 日经州长签署(州议会官网进程页)。HB 149 was signed by the Governor on June 22, 2025 (the Legislature's official history page).

已生效In force

2026 年 1 月 1 日。州议会官网进程页记 Effective on 1/1/26,州总检察长官网消费者 AI 权利页亦记 Effective January 1, 2026。January 1, 2026. The Legislature's official history page records "Effective on 1/1/26", and the Attorney General's consumer AI rights page likewise records "Effective January 1, 2026".

执法已动Enforcement

执法机制已建。州总检察长官网已设 TRAIGA 说明页与 AI 投诉入口,明示总检察长有专属执法权、可发民事调查要求、可诉请民事罚款;本次打开的官方页面未见已公布的处罚个案。The enforcement mechanism is in place. The Attorney General's website has a TRAIGA explainer page and an AI complaint portal, and states that the Attorney General has exclusive enforcement authority, may issue civil investigative demands and may sue for civil penalties. The official pages opened in this round showed no published penalty case.

待核Pending

是否有首例执法个案,待核。Whether a first enforcement case exists is pending verification.

实际关口Practical gate成法文本以禁止性条款为主,禁止以故意操纵诱导自伤、伤害他人或犯罪,禁止政府社会评分与政府以生物识别方式识别特定个人,禁止故意实施非法歧视;与消费者的 AI 交互披露义务只落在政府机构与医疗服务提供者,私营主体没有科州式的高风险影响评估义务。The enacted text is built mainly on prohibitions: no intentional manipulation inducing self-harm, harm to others or crime; no government social scoring and no government biometric identification of specific individuals; no intentional unlawful discrimination. The duty to disclose AI interaction to consumers falls only on government agencies and health care providers; private actors carry no Colorado-style high-risk impact-assessment duty.

R-06已生效 · 执法已动In force · enforcement active

加州 CCPA(经 CPRA 修订)California CCPA (as amended by the CPRA)

规则名Official nameCalifornia Consumer Privacy Act (CCPA),经 California Privacy Rights Act (CPRA, Proposition 24) 修订。引用时应写「CCPA(经 CPRA 修订)」,CPRA 未另立新法,官方通称仍为 CCPA 或 CCPA as amended。California Consumer Privacy Act (CCPA), as amended by the California Privacy Rights Act (CPRA, Proposition 24). Citations should read "CCPA (as amended by the CPRA)": the CPRA did not create a separate new statute, and official usage remains CCPA, or CCPA as amended.

法域Jurisdiction美国加利福尼亚州State of California, United States

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

2020 年经加州选民批准(加州隐私保护局 CPPA 官网明示 In 2020, California voters approved Proposition 24;州务卿官方选民指南收有 Prop 24 全文与分析)。Approved by California voters in 2020 (the California Privacy Protection Agency, CPPA, states on its website: "In 2020, California voters approved Proposition 24"; the Secretary of State's official voter guide carries the full text and analysis of Prop 24).

已生效In force

CPPA 官网明示 CPRA 对 CCPA 的修订于 2023 年 1 月 1 日生效(The CPRA amendments to the CCPA went into effect on January 1, 2023);其自动决策技术(ADMT)、隐私风险评估与网络安全审计配套法规已于 2026 年 1 月 1 日生效。The CPPA states that the CPRA amendments to the CCPA went into effect on January 1, 2023. Its supporting regulations on automated decision-making technology (ADMT), privacy risk assessments and cybersecurity audits took effect on January 1, 2026.

执法已动Enforcement

有。CPPA 官网明示该局负责实施与执法,可调查、审计并提起行政执法行动。Yes. The CPPA states that it is responsible for implementation and enforcement, and may investigate, audit and bring administrative enforcement actions.

待核Pending

具体执法个案清单与罚款额,未逐个打开 CPPA 个案页核,标待核。The list of individual enforcement cases and their fine amounts was not verified case by case against the CPPA's case pages, and is marked pending.

实际关口Practical gate达到门槛的企业处理加州居民个人信息时,须守目的限制与数据最小化,保障知情、删除、更正、退出出售或共享(含跨情境行为广告)与限制敏感个人信息使用等权利;AI 训练与画像所用个人信息同样受此约束。Businesses meeting the thresholds that process California residents' personal information must observe purpose limitation and data minimisation, and honour the rights to know, delete, correct, opt out of sale or sharing (including cross-context behavioural advertising) and limit the use of sensitive personal information. Personal information used for AI training and profiling is equally covered.

R-07已出台 · 义务自 2026-01-01 起算Promulgated · duties run from Jan 1, 2026

加州 AB 2013 训练数据透明California AB 2013, Training Data Transparency

规则名Official nameAB 2013, Generative Artificial Intelligence: Training Data Transparency(生成式人工智能训练数据透明法),新增民法典 Title 15.2 第 3110 至 3111 条AB 2013, Generative Artificial Intelligence: Training Data Transparency, adding Title 15.2, Sections 3110 to 3111, to the Civil Code

法域Jurisdiction美国加利福尼亚州State of California, United States

方向Direction出海Outbound

对应卡点EXP blocker架构与部署Architecture and deployment

已出台Promulgated

2024 年 9 月 28 日经州长批准并由州务卿存档,Chapter 817, Statutes of 2024(章本扉页文字已核;加州立法信息网法案状态页本次打开返回 403,未能直开,此缺口在此注明)。Approved by the Governor and filed with the Secretary of State on September 28, 2024, Chapter 817, Statutes of 2024 (the chapter title page was verified; the California Legislative Information bill status page returned a 403 in this round and could not be opened directly, a gap noted here).

已生效In force

条文义务自 2026 年 1 月 1 日起算,第 3111 条明示开发者须于 2026 年 1 月 1 日或之前完成首次公布。The duties in the text run from January 1, 2026; Section 3111 states expressly that a developer must complete its first publication on or before January 1, 2026.

执法已动Enforcement

条文本身未设专门罚则与执法机关条款;本次核查未见官方执法个案。The text itself sets no dedicated penalty or enforcement-authority provision, and this round's check found no official enforcement case.

待核Pending

执法路径(是否经不正当竞争法执行)原文无据,不写作定论;针对该法的诉讼进展未核官方文件,均标待核。The enforcement path (whether it runs through unfair competition law) has no basis in the text and is not written as a conclusion; litigation progress against the Act was not checked against official documents. Both are marked pending.

实际关口Practical gate2022 年 1 月 1 日或之后发布、向加州公众提供的生成式 AI 系统或服务,其开发者必须在自家网站公布训练数据文件(数据集来源与所有者、数据点数量与类型、是否含受版权等保护的数据、是否购买或许可取得、是否含个人信息等高层摘要),且每次实质性修改后、再次公开提供前须更新公布。A developer of a generative AI system or service released on or after January 1, 2022 and made available to Californians must publish training data documentation on its own website (a high-level summary covering dataset sources and owners, the number and types of data points, whether the data includes material protected by copyright or other rights, whether it was purchased or licensed, and whether it contains personal information), and must update the publication after each material change, before making the system available again.

官方来源Official sources

核查日期 2026-11-01Checked November 1, 2026

R-08已施行(2026-08-02)· SB 1000 修订已即时生效Operative (Aug 2, 2026) · SB 1000 amendments in immediate effect

加州 SB 942 AI 透明法California SB 942, AI Transparency Act

规则名Official nameSB 942, California AI Transparency Act(加州人工智能透明法),新增商业与职业法典第 25 章第 22757 条起SB 942, California AI Transparency Act, adding Chapter 25, commencing with Section 22757, to the Business and Professions Code

法域Jurisdiction美国加利福尼亚州State of California, United States

方向Direction出海Outbound

对应卡点EXP blocker架构与部署 / 分发渠道Architecture and deployment / Distribution channels

已出台Promulgated

2024 年 9 月 19 日经州长批准,Chapter 291, Statutes of 2024。其后 SB 1000 于 2026 年 9 月 30 日经州长批准并由州务卿存档,Chapter 861, Statutes of 2026,属紧急立法,自批准当日即时生效,修订第 22757.1 条至第 22757.5 条并新增后限期废止第 22757.1 条。Approved by the Governor on September 19, 2024, Chapter 291, Statutes of 2024. Afterwards, SB 1000 was approved by the Governor and filed with the Secretary of State on September 30, 2026, Chapter 861, Statutes of 2026; it is urgency legislation, effective immediately on approval, amending Sections 22757.1 to 22757.5 and newly providing for the later repeal of Section 22757.

已生效In force

原法第 22757.6 条定施行(operative)日期为 2026 年 1 月 1 日;其后 AB 853 于 2025 年 10 月 13 日经州长批准(Chapter 674),明示将该法施行日期推迟至 2026 年 8 月 2 日,并新增大型在线平台(2027 年 1 月 1 日起)与生成式 AI 托管平台义务。现行口径应记 2026 年 8 月 2 日,不得再单写 2026 年 1 月 1 日。SB 1000 未修订第 22757.6 条,2026 年 8 月 2 日施行日不变;SB 1000 的修订自 2026 年 9 月 30 日即时生效,覆盖范围与义务口径自该日起按修订后文本读。Section 22757.6 of the original Act set the operative date at January 1, 2026. Afterwards, AB 853, approved by the Governor on October 13, 2025 (Chapter 674), expressly delayed the operative date to August 2, 2026, and added duties for large online platforms (from January 1, 2027) and generative AI hosting platforms. The current reading is August 2, 2026; January 1, 2026 must no longer be written on its own. SB 1000 did not amend Section 22757.6, so the August 2, 2026 operative date stands; SB 1000's amendments took effect immediately on September 30, 2026, and coverage and duties are read under the amended text from that date.

执法已动Enforcement

经 SB 1000 修订的第 22757.4 条定每项违规民事罚款 5,000 美元、胜诉原告可获律师费,覆盖提供者、大型在线平台或拍摄设备制造商每持续一日计为一项独立违规,并新增豁免:本修订生效日前的行为、且在生效日后不违反本章的,先前诉讼不得维持。虚假声称系统主要功能为辅助技术的,不走第 22757.4 条,另按新增第 22757.1 条每项违规罚 50,000 美元,该条至 2029 年 1 月 1 日废止。本次打开的官方页面未见已公布的执法个案。Section 22757.4, as amended by SB 1000, sets a civil penalty of USD 5,000 per violation, with attorney's fees for a prevailing plaintiff; for a covered provider, a large online platform or a capture device manufacturer, each day of a continuing violation counts as a separate violation. It also adds a safe harbour: a prior action may not be maintained for conduct before the amendment's effective date that does not violate the chapter after that date. Falsely claiming that a system's primary function is assistive technology does not go through Section 22757.4; it is instead fined USD 50,000 per violation under the new Section 22757.1, which is repealed on January 1, 2029. The official pages opened in this round showed no published enforcement case.

待核Pending

无。原标待核的 SB 1000 修订(覆盖门槛、显性标识、验证工具、许可与执法条款)已按章本原文核清并入正文,待核解除。None. The SB 1000 amendments previously marked pending (coverage thresholds, manifest disclosure, verification tools, licensing and enforcement provisions) have been verified against the chaptered text and folded into this entry; the pending item is discharged.

实际关口Practical gate(按 SB 1000 修订后文本)创建、编写或以其他方式产出在加州境内可公开访问的生成式 AI 系统者即为覆盖提供者,SB 1000 已删除原「每月访客或用户超过 100 万」的门槛,覆盖范围扩大为无规模下限。覆盖提供者须免费提供披露验证工具(disclosure verification tool,取代原「AI 检测工具」称谓),使用户能核验图像、视频、音频或其组合内容是否由其系统生成或改动(轻微修改除外),工具须输出内容中检出的系统溯源数据,并收集用户对工具效用的反馈;工具可由合规的第三方工具替代。须在技术上可行的范围内,在其系统生成或改动的上述内容中嵌入隐性披露,直接或经永久网址链接载明:提供者名称、生成或改动该内容的系统名称与版本、生成或改动的时间与日期、唯一标识符、以及该系统系生成还是改动了该内容(2029 年 1 月 1 日起另须载明该系统是否主要设计为辅助技术);披露须为永久性或极难移除篡改、与披露验证工具兼容、并符合或可互通于广泛认可的行业标准。原法「向用户提供添加显性标识选项」的义务已经 SB 1000 删除,修订后文本不再有用户可选显性标识要求。许可条款同步重订:许可方在授权时须告知被许可方其本章义务;许可方明知可识别的被许可方改动被许可系统致其不再合规的,须在知悉后 72 小时内终止授权,或在 72 小时内通知被许可方其不合规及下述报告要求;被许可方须在收到通知后 96 小时内改至合规或停止使用,并向许可方报告所采取的行动;许可方未收到报告、或报告显示被许可方未改至合规也未停止使用的,须向总检察长报告该不合规,总检察长须建接收报告机制;许可方无法定监测调查义务。主要设计为辅助技术的生成式 AI 系统在 2029 年 1 月 1 日前不适用本章,覆盖提供者不得虚假声称系统属此情形。(Under the text as amended by SB 1000) Anyone who creates, codes or otherwise produces a generative AI system that is publicly accessible in California is a covered provider; SB 1000 deleted the original threshold of more than one million monthly visitors or users, so coverage now has no size floor. A covered provider must make available, free of charge, a disclosure verification tool (the term replaces the original "AI detection tool"), letting users verify whether image, video, audio or combined content was generated or altered by its system (minor modifications excepted); the tool must output the system provenance data detected in the content and collect user feedback on the tool's effectiveness, and a compliant third-party tool may substitute. The provider must, to the extent technically feasible, embed a latent disclosure in such content its system generates or alters, stating directly or through a link to a permanent URL: the provider's name; the name and version of the system that generated or altered the content; the time and date of the generation or alteration; a unique identifier; and whether the system generated or altered the content (from January 1, 2029, also whether the system is primarily designed as assistive technology). The disclosure must be permanent or extremely difficult to remove or alter, compatible with the disclosure verification tool, and conform to, or be interoperable with, widely recognised industry standards. The original Act's duty to offer users an option to add a manifest disclosure has been deleted by SB 1000; the amended text no longer requires a user-selectable manifest disclosure. The licensing provisions were rewritten in step: a licensor must inform the licensee of its duties under the chapter at the time of licensing; if the licensor knowingly identifies that a licensee has modified the licensed system so it is no longer compliant, it must terminate the licence within 72 hours of learning this, or notify the licensee within 72 hours of the non-compliance and of the reporting requirement below; the licensee must, within 96 hours of receiving notice, bring the system into compliance or cease use, and report the action taken to the licensor; if the licensor receives no report, or the report shows the licensee neither complied nor ceased use, it must report the non-compliance to the Attorney General, who must build a mechanism to receive such reports; the licensor has no statutory duty to monitor or investigate. Generative AI systems primarily designed as assistive technology are outside the chapter until January 1, 2029, and a covered provider must not falsely claim that its system is of that kind.

R-09现行有效 · 执法已动且口径有回调In force · enforcement active, posture partly rolled back

FTC 法第五条FTC Act Section 5

规则名Official nameSection 5 of the Federal Trade Commission Act, 15 U. § 45(联邦贸易委员会法第五条,不公平或欺骗性行为禁止条款)Section 5 of the Federal Trade Commission Act, 15 U. § 45 (the prohibition on unfair or deceptive acts or practices)

法域Jurisdiction美国联邦United States, federal

方向Direction出海Outbound

对应卡点EXP blocker监管审批(作 AI 执法依据记)Regulatory approval (recorded as an AI enforcement basis)

已出台Promulgated

该条长期为现行法,编入 15 U. § 45;原始制定年份未打开 FTC 官网史页核,标待核。本条以 FTC 官网 2025 年 7 月修订的执法权限备忘为核查依据。The provision has long been current law, codified at 15 U. § 45; its original year of enactment was not checked against the FTC's official history page and is marked pending. This entry rests on the FTC's enforcement-authority memorandum, revised July 2025, on its website.

已生效In force

现行有效。备忘引第 5(a) 条原文,unfair or deceptive acts or practices in or affecting commerce are declared unlawful;并引第 45(n) 条对 unfair 的三要件定义(造成或可能造成消费者重大损害、消费者自身无法合理避免、且不被对消费者或竞争的抵消性利益所超过)。Currently in force. The memorandum quotes Section 5(a): "unfair or deceptive acts or practices in or affecting commerce are declared unlawful"; and quotes the three-part definition of "unfair" in Section 45(n) (causes or is likely to cause substantial injury to consumers, which is not reasonably avoidable by consumers themselves, and not outweighed by countervailing benefits to consumers or to competition).

执法已动Enforcement

有,且口径有回调。FTC 于 2024 年 9 月 25 日宣布 Operation AI Comply 执法行动(DoNotPay、Rytr、Ascend Ecom 等),官网原话为 Using AI tools to trick, mislead, or defraud people is illegal,与 there is no AI exemption from the laws on the books;同时 FTC 官网 2025 年文件显示其已重开并撤销 Rytr 同意令,理由是原投诉不足以支持第五条违反,说明 2025 年后的执法口径较 2024 年有收缩,引用时两面都要记。Yes, and the posture has partly rolled back. On September 25, 2024 the FTC announced its Operation AI Comply enforcement sweep (DoNotPay, Rytr, Ascend Ecom and others), stating on its website that "Using AI tools to trick, mislead, or defraud people is illegal" and that "there is no AI exemption from the laws on the books". At the same time, FTC documents from 2025 show that it reopened and set aside the Rytr consent order, on the ground that the original complaint could not support a Section 5 violation, which shows the post-2025 enforcement posture is narrower than in 2024; citations must record both sides.

待核Pending

不得把 2024 年的专案表述直接当作现行统一口径;单个 AI 案件是否仍按第五条推进,须以 FTC 官网个案页逐个核。The 2024 sweep's framing must not be taken directly as the current uniform posture; whether an individual AI case still proceeds under Section 5 must be checked case by case on the FTC's case pages.

实际关口Practical gate没有 AI 豁免,也没有 AI 专项上市前审批;AI 能力与效果宣称若构成对合理消费者的重大误导,或部署方式造成不可合理避免的重大损害,即落入第五条执法范围,宣称须有证据支撑。There is no AI exemption and no AI-specific pre-market approval. If claims about AI capability or performance materially mislead reasonable consumers, or a deployment causes substantial injury that consumers cannot reasonably avoid, the conduct falls inside Section 5 enforcement, and claims must be backed by evidence.

R-10已生效In force

韩国个人信息保护法(PIPA)Korea's Personal Information Protection Act (PIPA)

规则名Official name개인정보 보호법 (Personal Information Protection Act),中文通称韩国《个人信息保护法》개인정보 보호법 (Personal Information Protection Act)

法域Jurisdiction韩国South Korea

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

2011 年 3 月 29 日公布,法律第 10465 号(国家法令信息中心 law.go.kr 原文标注)。Promulgated on March 29, 2011, as Act No. 10465 (as annotated in the text at the National Law Information Center, law.go.kr).

已生效In force

2011 年 9 月 30 日施行(原文附则:公布后 6 个月施行)。此后多次修订;law.go.kr 现行页另有 2026 年 3 月 10 日部分修订(法律第 21445 号,2026 年 9 月 11 日施行版)与更晚修订并存,引用必须注明版本日期。In force from September 30, 2011 (the Addenda set effect six months after promulgation). It has been amended many times since; the current law.go.kr page carries, side by side, a partial amendment of March 10, 2026 (Act No. 21445, the version in force from September 11, 2026) and later amendments, so citations must state the version date relied on.

执法已动Enforcement

主管机关为个人信息保护委员会(PIPC);本次未逐个打开 PIPC 处罚个案页核,个案标待核。The competent authority is the Personal Information Protection Commission (PIPC). This round did not open the PIPC's individual penalty case pages one by one; the cases are marked pending.

待核Pending

现行文本版本滚动较快,2027 年施行的一批修订条款与现行条文的对应关系待逐条核。The current text rolls forward quickly; how the batch of amendments taking effect in 2027 maps onto the current provisions remains to be checked article by article.

实际关口Practical gate与 AI 直接相关的硬条款是第 37 条之 2(PIPC 官网指南明示自 2024 年 3 月 15 日起施行):以包括 AI 系统在内的完全自动化系统处理个人信息、作出对信息主体权利义务有重大影响的决定时,信息主体有权拒绝该决定并要求说明,处理者须公开自动化决定的基准、程序与个人信息处理方式,无正当理由须不予适用或经人工介入重新处理。The hard provision directly touching AI is Article 37-2 (the PIPC's official guidance states it has been in force since March 15, 2024): where personal information is processed by a fully automated system, including an AI system, to make a decision with a significant effect on the data subject's rights or obligations, the data subject may refuse the decision and demand an explanation; the processor must disclose the criteria and procedures of the automated decision and how the personal information is processed, and, absent justifiable grounds, must not apply the decision or must re-process it with human involvement.

R-11已生效(2026-01-22)· 罚款设宽限期In force (Jan 22, 2026) · fine grace period

韩国《人工智能基本法》Korea's Framework Act on Artificial Intelligence

规则名Official name인공지능 발전과 신뢰 기반 조성 등에 관한 기본법,中文通称韩国《人工智能基本法》인공지능 발전과 신뢰 기반 조성 등에 관한 기본법 (Framework Act on the Development of Artificial Intelligence and Establishment of a Foundation for Trust)

法域Jurisdiction韩国South Korea

方向Direction出海Outbound

对应卡点EXP blocker监管审批 / 架构与部署Regulatory approval / Architecture and deployment

已出台Promulgated

通过与公布分开记。国会全体会议通过为 2024 年 12 月 26 日(科学技术信息通信部官网新闻稿明示);公布为 2025 年 1 月 21 日,法律第 20676 号(law.go.kr 原文标注 2025. 21. 제정)。Passage and promulgation are recorded separately. The National Assembly passed the Act in plenary on December 26, 2024 (stated in a press release on the Ministry of Science and ICT website); it was promulgated on January 21, 2025, as Act No. 20676 (annotated in the law.go.kr text as enacted on 2025. 21).

已生效In force

2026 年 1 月 22 日(law.go.kr 原文标注 시행 2026. 22.,附则为公布后满 1 年施行;其中数字医疗器械相关部分为 2026 年 1 月 24 日)。January 22, 2026 (annotated in the law.go.kr text as in force from 2026. 22; the Addenda set effect one year after promulgation; the parts concerning digital medical devices are dated January 24, 2026).

执法已动Enforcement

韩国政府公开口径为施行初期设至少 1 年的罚款指导宽限期,以引导合规为主;本次未见官方公布的处罚个案,标待核。The Korean government's stated position is a guidance grace period of at least one year on fines in the early period after effect, aimed at guiding compliance. This round found no officially published penalty case; marked pending.

待核Pending

2026 年 1 月 20 日部分修订(法律第 21311 号)新增条款的分批施行日期须逐条对 law.go.kr 附则核,本次未逐条核完。The staged effective dates of the provisions added by the partial amendment of January 20, 2026 (Act No. 21311) must be checked article by article against the law.go.kr Addenda; this round did not finish that check.

实际关口Practical gate第 31 条透明度义务为硬关口,提供高影响 AI 或生成式 AI 产品或服务须事前告知用户其基于 AI 运行,生成式 AI 产出物须标示为 AI 生成,难以与真实区分的虚拟音像产出须以用户能明确识别的方式告知或标示;第 33 条要求经营者在提供产品或服务前先行检视其 AI 是否属高影响 AI,必要时可请求科学技术信息通信部长官确认。The Article 31 transparency duties are the hard gate: a business offering a high-impact AI or generative AI product or service must tell users in advance that it runs on AI; generative AI outputs must be marked as AI-generated; and virtual audio or visual outputs that are hard to distinguish from real ones must be notified or marked in a way users can clearly recognise. Article 33 requires a business, before offering a product or service, to examine whether its AI is high-impact AI, and allows it to ask the Minister of Science and ICT for confirmation where needed.

R-12已生效 · 滚动修订中In force · under rolling amendment

日本个人信息保护法(APPI)Japan's Act on the Protection of Personal Information (APPI)

规则名Official name個人情報の保護に関する法律 (Act on the Protection of Personal Information),平成 15 年法律第 57 号,中文通称日本《个人信息保护法》個人情報の保護に関する法律 (Act on the Protection of Personal Information), Act No. 57 of 2003

法域Jurisdiction日本Japan

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

个人情报保护委员会(PPC)官网沿革记 2003 年 5 月 23 日成立、2003 年 5 月 30 日公布并部分施行,法律编号为平成 15 年 5 月 30 日法律第 57 号。The Personal Information Protection Commission (PPC) records in its official history that the Act was established on May 23, 2003, promulgated and partly brought into force on May 30, 2003, under Act No. 57 of May 30, 2003.

已生效In force

2005 年 4 月 1 日全面施行(PPC 官网沿革)。其后 2020 年修订法于 2020 年 6 月 5 日国会通过成立、6 月 12 日公布,全面施行日为 2022 年 4 月 1 日(PPC 官网修订页明示)。Fully in force from April 1, 2005 (PPC official history). The later 2020 amendment act was passed by the Diet on June 5, 2020 and promulgated on June 12, 2020, with full effect from April 1, 2022 (stated on the PPC's amendment page).

执法已动Enforcement

PPC 官网文件明示该会统一监管民间与公共部门并执行本法;本次未逐个打开 PPC 命令或劝告个案页核,个案标待核。PPC documents state that the Commission supervises both the private and public sectors uniformly and enforces the Act. This round did not open the PPC's individual order or recommendation case pages one by one; the cases are marked pending.

待核Pending

PPC 法令页已挂出更晚施行版本的对照表,本法仍在滚动修订,引用条文须注明所据版本日期。The PPC's legislation page already carries comparison tables for versions with later effective dates; the Act is still under rolling amendment, and citations must state the version date relied on.

实际关口Practical gate本法不是 AI 专项法,关口在数据侧:处理个人信息的经营者须守利用目的特定、正当取得、安全管理与第三方提供(含向外国第三者提供,PPC 设有专门指南)等义务;AI 训练与服务所用个人数据在日本境内处理或向境外提供时须先过这一关。This is not an AI-specific statute; the gate is on the data side. A business handling personal information must observe the duties on specifying the purpose of use, proper acquisition, security controls and provision to third parties (including provision to third parties in foreign countries, for which the PPC has a dedicated guideline). Personal data used in AI training and services must clear this gate when it is processed in Japan or provided abroad.

R-13已生效(分段)In force (in stages)

巴西通用个人数据保护法(LGPD)Brazil's General Personal Data Protection Law (LGPD)

规则名Official nameLei nº 13.709, de 14 de agosto de 2018, Lei Geral de Proteção de Dados Pessoais (LGPD),中文通称巴西《通用个人数据保护法》Lei nº 13.709, de 14 de agosto de 2018, Lei Geral de Proteção de Dados Pessoais (LGPD)

法域Jurisdiction巴西(联邦法律)Brazil (federal law)

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

2018 年 8 月 14 日签署,公布于 2018 年 8 月 15 日官方公报 DOU 并于同日增刊部分重刊。Lei nº 13.853, de 2019 已把短标题定为 Lei Geral de Proteção de Dados Pessoais (LGPD) 并设数据保护局 ANPD。Signed on August 14, 2018, published in the official gazette (DOU) on August 15, 2018 and partly republished in an extra edition the same day. Lei nº 13.853, de 2019 set the short title as Lei Geral de Proteção de Dados Pessoais (LGPD) and created the data protection authority, the ANPD.

已生效In force

分段生效,以汇编本 Article 65 为准逐个列:第 I 项,ANPD 相关条文(Articles 55-A 至 55-L、58-A、58-B)自 2018 年 12 月 28 日生效;第 I-A 项(由 Lei nº 14.010, de 2020 加入),制裁条文 Articles 52、53、54 自 2021 年 8 月 1 日生效;第 II 项,其余条文为公布后 24 个月生效。In force in stages, listed one by one from Article 65 of the consolidated text: item I, the ANPD provisions (Articles 55-A to 55-L, 58-A and 58-B) took effect on December 28, 2018; item I-A (added by Lei nº 14.010, de 2020), the sanctions provisions, Articles 52, 53 and 54, took effect on August 1, 2021; item II, the remaining provisions took effect 24 months after publication.

执法已动Enforcement

有机制且机构已设。Article 55-A 设 Autoridade Nacional de Proteção de Dados (ANPD);Article 52 定简单罚款为巴西境内营业额 2%,每项违规上限 5000 万雷亚尔,另有警告、封存与删除数据等制裁。The mechanism exists and the authority is established. Article 55-A creates the Autoridade Nacional de Proteção de Dados (ANPD); Article 52 sets the simple fine at 2% of revenue in Brazil, capped at BRL 50 million per violation, alongside warnings and the blocking and deletion of data among other sanctions.

待核Pending

其余条文的实际操作生效日(2020 年 9 月 18 日一说)未能仅凭汇编本 Article 65 干净读出,汇编本同页并列了 Medida Provisória nº 959, de 2020 曾改写第 II 项的版本,该日后经 Lei nº 14.010, de 2020 落定的口径须再以 DOU 最终文本逐字复核,本项暂记待核,不把 2020 年 9 月 18 日写成已核定论。另注:Planalto 原页本次直开遭拦截,条文系经 gov.br 官方汇编 PDF 逐条核读,原页链接仍列 Planalto。The operative effective date of the remaining provisions (one account says September 18, 2020) cannot be read cleanly from Article 65 of the consolidated text alone: the same page carries, side by side, the version of item II once rewritten by Medida Provisória nº 959, de 2020, and the reading finally settled by Lei nº 14.010, de 2020 must be re-verified word by word against the final DOU text. This point is recorded as pending, and September 18, 2020 is not written as a verified conclusion. Note also: the Planalto original page was blocked when opened directly in this round; the provisions were read and checked article by article from the official gov.br consolidated PDF, while the source link still points to Planalto.

实际关口Practical gate原文 Article 3 定,处理在巴西境内进行,或以向巴西境内个人提供商品服务或处理其数据为目的,或数据在巴西境内收集的,均适用本法;AI 产品在巴西处理个人数据须落入 Article 7 等合法基础之一。Article 3 of the text provides that the Law applies where the processing takes place in Brazil, or is aimed at offering goods or services to individuals in Brazil or processing their data, or where the data was collected in Brazil. An AI product processing personal data in Brazil must rest on one of the lawful bases, such as those in Article 7.

R-14分三段指定生效 · 核心义务在第三段Commencement in three tranches · core duties in the third

印度数字个人数据保护法(DPDP Act)India's Digital Personal Data Protection Act (DPDP Act)

规则名Official nameThe Digital Personal Data Protection Act, 2023 (No. 22 of 2023),中文通称印度《数字个人数据保护法》The Digital Personal Data Protection Act, 2023 (No. 22 of 2023)

法域Jurisdiction印度(议会法案,主管为电子与信息技术部 MeitY)India (an Act of Parliament; administered by the Ministry of Electronics and Information Technology, MeitY)

方向Direction出海Outbound

对应卡点EXP blocker数据与隐私合规Data and privacy compliance

已出台Promulgated

2023 年 8 月 11 日获总统同意并刊于印度官方公报 The Gazette of India, Extraordinary, Part II, Section 1, No. 25,公报号 CG-DL-E-12082023-248045。Received the President's assent on August 11, 2023 and was published in The Gazette of India, Extraordinary, Part II, Section 1, No. 25, gazette no. CG-DL-E-12082023-248045.

已生效In force

法案本身未定统一生效日。Section 1(2) 原文定由中央政府以官方公报通知指定生效日,不同条文可定不同日期。其后 MeitY 以 G. 843(E)(通知日 2025 年 11 月 13 日,公报原件已开核)分三段指定:第一段公布日即生效,含 Section 1(2)、Section 2、Sections 18 至 26(Data Protection Board 设立)、Sections 35、38 至 43、Section 44(1) 与 (3);第二段公布后一年生效,即 Section 6(9) 与 Section 27(1)(d);第三段公布后十八个月生效,含 Sections 3 至 5、Section 6(1) 至 (8) 与 (10)、Sections 7 至 17、Section 27 除 (1)(d) 外、Sections 28 至 34、Sections 36、37 与 Section 44(2),即数据受托人核心义务与罚则程序所在段。The Act itself sets no single commencement date. Section 1(2) provides that the Central Government designates the commencement date by notification in the Official Gazette, and different dates may be set for different provisions. MeitY then designated three tranches by G. 843(E) (notification dated November 13, 2025; the gazette original was opened and checked): the first tranche took effect on the day of publication, covering Section 1(2), Section 2, Sections 18 to 26 (establishing the Data Protection Board), Sections 35 and 38 to 43, and Section 44(1) and (3); the second tranche takes effect one year after publication, covering Section 6(9) and Section 27(1)(d); the third tranche takes effect eighteen months after publication, covering Sections 3 to 5, Section 6(1) to (8) and (10), Sections 7 to 17, Section 27 except (1)(d), Sections 28 to 34, Sections 36 and 37, and Section 44(2), the tranche carrying the data fiduciaries' core obligations and the penalty procedures.

执法已动Enforcement

尚属机构设立阶段,核心执法未动。Board 相关条文已生效,但主体义务与罚则条文在第三段才生效,该段到期前不能写成全面执法已动。Still at the institution-building stage; core enforcement has not moved. The Board provisions are in force, but the substantive duties and the penalty provisions take effect only in the third tranche; before that tranche falls due, this must not be written as full enforcement in motion.

待核Pending

一年段与十八个月段的具体日历日待核。G. 843(E) 原件只写公布后一年与公布后十八个月,公报刊期日为 2025 年 11 月 13 日而公报文件号含 14112025,二级汇总因此有数到 2027 年 5 月 13 日与 2027 年 5 月 14 日两说,须以 MeitY 对公布日的正式口径确认后才可写死。法案附表各项罚金上限本次未在 MeitY 法案原文页逐项核到,不报具体金额。The exact calendar dates of the one-year and eighteen-month tranches are pending. The G. 843(E) original states only one year and eighteen months after publication; the gazette issue date is November 13, 2025 while the gazette document number contains 14112025, so secondary summaries count to both May 13, 2027 and May 14, 2027. The dates can only be fixed once MeitY's official reading of the publication date is confirmed. The penalty ceilings in the Act's Schedule were not verified item by item on MeitY's Act page in this round, and no figures are reported.

实际关口Practical gate适用于数字个人数据的处理,数据受托人须以同意或法定用途为基础处理并履行告知、安全保障与泄露通知义务;但这些核心义务条文按 G. 843(E) 在十八个月段才生效,AI 产品入印的硬合规截止随该段走。It applies to the processing of digital personal data: data fiduciaries must process on the basis of consent or a legitimate use, and must meet notice, security-safeguard and breach-notification duties. But under G. 843(E) those core duty provisions take effect only in the eighteen-month tranche, and the hard compliance deadline for an AI product entering India follows that tranche.

二、中国出境门Part II · China's outbound gate

R-15已生效 · 阈值以 R-17 现行口径为准In force · thresholds read under R-17

数据出境安全评估办法Measures for Data Export Security Assessment

规则名Official name《数据出境安全评估办法》,国家互联网信息办公室令第 11 号Measures for Data Export Security Assessment, Order No. 11 of the Cyberspace Administration of China

法域Jurisdiction中国China

方向Direction出境门Outbound gate

对应卡点EXP blocker数据与隐私合规 / 监管审批Data and privacy compliance / Regulatory approval

已出台Promulgated

2022 年 7 月 7 日公布。令文载明已经 2022 年 5 月 19 日国家互联网信息办公室 2022 年第 10 次室务会议审议通过,令文落款 2022 年 7 月 7 日。Published on July 7, 2022. The order records that the Measures were deliberated and adopted at the Cyberspace Administration of China's 10th office meeting of 2022, on May 19, 2022, and the order is dated July 7, 2022.

已生效In force

2022 年 9 月 1 日起施行,见令文及第二十条。第二十条另规定施行前已开展的不符合规定的数据出境活动,应当自施行之日起 6 个月内完成整改。In force from September 1, 2022, per the order and Article 20. Article 20 further provides that data export activities already under way before implementation that do not conform must be rectified within six months of the implementation date.

执法已动Enforcement

有。国家网信办 2026 年 3 月 23 日官方回顾称,已组织开展重点企业机构数据出境管理专项检查,对违法违规问题依法依规进行处置处罚,并已发布申报指南第三版与数据出境申报系统,评估申报在实际运行。Yes. The Cyberspace Administration of China's official review of March 23, 2026 states that it has organised special inspections of data export management at key enterprises and institutions, handled and penalised violations in accordance with law and regulation, and has issued the third edition of the filing guide and the data export filing system; assessment filings are operating in practice.

待核Pending

本办法原文自身无待核日期项;但第四条原始阈值已被 R-17 调整,现行适用口径以 R-17 为准,不得只引本办法原文判断是否需申报。The Measures themselves carry no pending date item; but the original Article 4 thresholds have been adjusted by R-17, and the currently applicable reading follows R-17. Whether a filing is required must not be judged from these Measures alone.

原文阈值(原始口径,仅作历史基准)Original thresholds (original wording, historical baseline only)

第四条规定,下列情形之一应当申报数据出境安全评估:一是向境外提供重要数据;二是关键信息基础设施运营者和处理 100 万人以上个人信息的数据处理者向境外提供个人信息;三是自上年 1 月 1 日起累计向境外提供 10 万人个人信息或者 1 万人敏感个人信息;四是国家网信部门规定的其他情形。第十四条原始口径为评估结果有效期 2 年,该有效期后被 R-17 改为 3 年。Article 4 provides that a data export security assessment must be filed in any of these cases: (1) important data is provided abroad; (2) a critical information infrastructure operator, or a data processor handling the personal information of more than one million people, provides personal information abroad; (3) since January 1 of the previous year, the cumulative total provided abroad reaches the personal information of 100,000 people or the sensitive personal information of 10,000 people; (4) other cases specified by the national cyberspace authority. The original reading of Article 14 set the validity of an assessment result at two years; R-17 later changed that to three years.

实际关口Practical gateAI 产品在境内运营中收集和产生的重要数据和个人信息向境外提供时,落入上述情形(按现行口径读 R-17)的须先经省级网信部门向国家网信部门申报安全评估,未通过评估不得开展所申报的数据出境活动,原文依据为第二条、第四条及答记者问口径。Where an AI product provides abroad important data or personal information collected and generated in its domestic operations, and the case falls within the situations above (read under the current wording, R-17), a security assessment must first be filed through the provincial cyberspace authority to the national authority, and the declared data export activity may not proceed without passing the assessment. The textual basis is Articles 2 and 4 and the official press-briefing reading.

R-16已生效 · 阈值以 R-17 现行口径为准In force · thresholds read under R-17

个人信息出境标准合同办法Measures for the Standard Contract for Personal Information Export

规则名Official name《个人信息出境标准合同办法》,国家互联网信息办公室令第 13 号Measures for the Standard Contract for Personal Information Export, Order No. 13 of the Cyberspace Administration of China

法域Jurisdiction中国China

方向Direction出境门Outbound gate

对应卡点EXP blocker数据与隐私合规 / 监管审批Data and privacy compliance / Regulatory approval

已出台Promulgated

公布日有两种表述并存,令文落款为 2023 年 2 月 22 日,R-17 第十三条亦称其为 2023 年 2 月 22 日公布;中国网信网该页刊载时间为 2023 年 2 月 24 日,公布稿同日称 2 月 24 日公布。引用时应注明口径。令文载明已经 2023 年 2 月 3 日国家互联网信息办公室 2023 年第 2 次室务会议审议通过。Two datings of publication coexist: the order is dated February 22, 2023, and Article 13 of R-17 also calls it published on February 22, 2023; the CAC website page carries a posting time of February 24, 2023, and the promulgation release of the same day calls it published on February 24. Citations should state which dating is used. The order records deliberation and adoption at the Administration's 2nd office meeting of 2023, on February 3, 2023.

已生效In force

2023 年 6 月 1 日起施行,见令文及第十三条。第十三条另规定施行前已开展的不符合规定的个人信息出境活动,应当自施行之日起 6 个月内完成整改。In force from June 1, 2023, per the order and Article 13. Article 13 further provides that personal information export activities already under way that do not conform must be rectified within six months of the implementation date.

执法已动Enforcement

有。国家网信办官方回顾称各地已开展个人信息出境标准合同备案工作,广东已编制发布粤港澳大湾区个人信息跨境流动标准合同备案指南,国家网信办已开通数据出境申报系统受理备案,并对数据出境违法违规问题开展专项检查与处置处罚。Yes. The CAC's official review states that localities have carried out standard-contract filing work for personal information export; that Guangdong has compiled and issued a filing guide for the Guangdong to Hong Kong to Macao Greater Bay Area standard contract for cross-border personal information flows; that the CAC has opened the data export filing system to accept filings; and that special inspections, handling and penalties for data export violations have been carried out.

待核Pending

公布日双口径如上,分开记。本办法第四条原始适用阈值已被 R-17 调整,现行判断以 R-17 第五条、第七条、第八条为准。The two publication datings above are recorded separately. The original Article 4 applicability thresholds of these Measures have been adjusted by R-17; current determinations follow Articles 5, 7 and 8 of R-17.

原文阈值(原始口径,仅作历史基准)Original thresholds (original wording, historical baseline only)

第四条规定,通过订立标准合同向境外提供个人信息应当同时符合四项情形:一是非关键信息基础设施运营者;二是处理个人信息不满 100 万人;三是自上年 1 月 1 日起累计向境外提供个人信息不满 10 万人;四是自上年 1 月 1 日起累计向境外提供敏感个人信息不满 1 万人。并禁止以数量拆分等手段规避应走安全评估的情形。第六条规定标准合同生效后方可开展出境活动,第七条规定应当在标准合同生效之日起 10 个工作日内向所在地省级网信部门备案,并先开展个人信息保护影响评估,见第五条。Article 4 provides that providing personal information abroad by concluding a standard contract requires all four conditions at once: (1) the provider is not a critical information infrastructure operator; (2) it processes the personal information of fewer than one million people; (3) since January 1 of the previous year it has cumulatively provided abroad the personal information of fewer than 100,000 people; (4) since January 1 of the previous year it has cumulatively provided abroad the sensitive personal information of fewer than 10,000 people. Splitting volumes or similar means to evade a required security assessment is prohibited. Article 6 provides that the export activity may proceed only after the standard contract takes effect; Article 7 requires filing with the provincial cyberspace authority at the provider's location within ten working days of the contract taking effect; and a personal information protection impact assessment must be carried out first, per Article 5.

实际关口Practical gateAI 产品以订立标准合同方式向境外提供个人信息的,须先做个人信息保护影响评估、按附件订立标准合同、合同生效后方可出境并在 10 个工作日内备案,原文依据为第二条、第五条、第六条、第七条。An AI product that provides personal information abroad by the standard-contract route must first carry out a personal information protection impact assessment, conclude the standard contract in the annexed form, and may export only after the contract takes effect, filing within ten working days. The textual basis is Articles 2, 5, 6 and 7.

R-17已生效 · 现行分档口径以本规定为准In force · the current tiering follows this instrument

促进和规范数据跨境流动规定Provisions on Promoting and Regulating Cross-Border Data Flows

规则名Official name《促进和规范数据跨境流动规定》,国家互联网信息办公室令第 16 号Provisions on Promoting and Regulating Cross-Border Data Flows, Order No. 16 of the Cyberspace Administration of China

法域Jurisdiction中国China

方向Direction出境门Outbound gate

对应卡点EXP blocker数据与隐私合规 / 监管审批Data and privacy compliance / Regulatory approval

已出台Promulgated

2024 年 3 月 22 日公布。令文载明已经 2023 年 11 月 28 日国家互联网信息办公室 2023 年第 26 次室务会议审议通过,令文落款 2024 年 3 月 22 日。Published on March 22, 2024. The order records deliberation and adoption at the Administration's 26th office meeting of 2023, on November 28, 2023, and the order is dated March 22, 2024.

已生效In force

自公布之日起施行,见令文及第十四条,即 2024 年 3 月 22 日起施行。In force from the date of publication, per the order and Article 14, that is, from March 22, 2024.

执法已动Enforcement

有。国家网信办官方回顾直接以本规定实施两周年为题,称其出台实施以来已放宽数据跨境流动条件、收窄数据出境安全评估范围,并已落地自贸试验区负面清单、预评估试点与专项检查处置。Yes. The CAC's official review, written for the second anniversary of these Provisions, states that since their issue and implementation the conditions for cross-border data flows have been relaxed and the scope of data export security assessment narrowed, and that pilot free trade zone negative lists, pre-assessment pilots and special-inspection handling have been put in place.

待核Pending

第十三条规定,R-15 与 R-16 等相关规定与本规定不一致的,适用本规定,故前两项的现行阈值与有效期必须按本规定改读,无需待核。自由贸易试验区负面清单为各区分别制定与备案,具体清单内容不在本规定原文内,属逐区待核。Article 13 provides that where R-15, R-16 and other related provisions are inconsistent with these Provisions, these Provisions apply; the current thresholds and validity periods of the two earlier instruments must therefore be re-read under this one, and nothing there needs to stay pending. The pilot free trade zone negative lists are drawn up and filed zone by zone; their specific contents are not in the text of these Provisions and are pending zone by zone.

对前两项阈值的调整要点(只记原文)How it adjusts the two earlier instruments' thresholds (original text only)

重要数据口径,第二条定,未被相关部门、地区告知或者公开发布为重要数据的,数据处理者不需要作为重要数据申报数据出境安全评估。免予三项程序的情形,第三条至第五条:国际贸易、跨境运输、学术合作、跨国生产制造和市场营销等活动中收集和产生的数据向境外提供,不包含个人信息或者重要数据的,免予申报安全评估、订立标准合同、通过个人信息保护认证;在境外收集和产生的个人信息传输至境内处理后向境外提供且未引入境内个人信息或者重要数据的,同上免予;为订立或履行个人作为一方当事人的合同确需提供、按劳动规章与集体合同实施跨境人力资源管理确需提供员工个人信息、紧急情况下为保护自然人生命健康和财产安全确需提供的,免予;关键信息基础设施运营者以外的数据处理者自当年 1 月 1 日起累计向境外提供不满 10 万人个人信息(不含敏感个人信息)的,免予。第五条末款注明所称个人信息不包括重要数据。Important data, per Article 2: where the relevant departments or regions have not notified the processor, or publicly released, that data is important data, the processor need not file a data export security assessment treating it as important data. Exemptions from the three procedures, Articles 3 to 5: data collected and generated in activities such as international trade, cross-border transport, academic cooperation, transnational production and manufacturing, and marketing, provided abroad and containing no personal information or important data, is exempt from the security assessment, the standard contract and personal information protection certification; personal information collected and generated abroad, transferred into China for processing and then provided abroad again, without introducing domestic personal information or important data, is likewise exempt; provision genuinely needed to conclude or perform a contract to which the individual is a party, provision of employee personal information genuinely needed for cross-border human resources management under labour rules and collective contracts, and provision genuinely needed in an emergency to protect a natural person's life, health or property safety, are exempt; a data processor other than a critical information infrastructure operator that, since January 1 of the current year, has cumulatively provided abroad the personal information of fewer than 100,000 people (excluding sensitive personal information) is exempt. The last paragraph of Article 5 notes that the personal information referred to does not include important data.

应申报安全评估的现行阈值,第七条:关键信息基础设施运营者向境外提供个人信息或者重要数据;关键信息基础设施运营者以外的数据处理者向境外提供重要数据,或者自当年 1 月 1 日起累计向境外提供 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。计数起点由前两项原文的上年 1 月 1 日起改为当年 1 月 1 日起。应订立标准合同或者通过认证的现行阈值,第八条:关键信息基础设施运营者以外的数据处理者自当年 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 万人敏感个人信息的,应当依法订立个人信息出境标准合同或者通过个人信息保护认证。有效期,第九条:通过数据出境安全评估的结果有效期改为 3 年,届满前符合条件可申请延长,经国家网信部门批准可延长 3 年,取代 R-15 第十四条的 2 年口径。自贸区负面清单,第六条:自贸试验区可自行制定负面清单,经省级网络安全和信息化委员会批准后报国家网信部门、国家数据管理部门备案,区内向境外提供负面清单外的数据可免予三项程序。The current security-assessment thresholds, Article 7: a critical information infrastructure operator providing personal information or important data abroad; or a data processor other than such an operator providing important data abroad, or, since January 1 of the current year, cumulatively providing abroad the personal information of one million or more people (excluding sensitive personal information) or the sensitive personal information of 10,000 or more people. The counting start moved from January 1 of the previous year, as the two earlier instruments had it, to January 1 of the current year. The current standard-contract or certification thresholds, Article 8: a data processor other than a critical information infrastructure operator that, since January 1 of the current year, has cumulatively provided abroad the personal information of 100,000 or more but fewer than one million people (excluding sensitive personal information), or the sensitive personal information of fewer than 10,000 people, must conclude a personal information export standard contract or pass personal information protection certification, in accordance with law. Validity, Article 9: the validity of a passed data export security assessment is now three years; an extension may be applied for before expiry where the conditions are met, and, with the national cyberspace authority's approval, may be extended by three years, replacing the two-year reading at Article 14 of R-15. Free trade zone negative lists, Article 6: a pilot free trade zone may draw up its own negative list, which, after approval by the provincial cybersecurity and informatization commission, is filed with the national cyberspace authority and the national data authority; data outside the list provided abroad from within the zone may be exempt from the three procedures.

实际关口Practical gateAI 产品数据出境现按本规定的分档判断,不满 10 万人非敏感个人信息可免予三项程序,10 万人以上不满 100 万人档走标准合同或者认证,100 万人以上或敏感个人信息 1 万人以上或重要数据档须申报安全评估,原文依据为第五条、第七条、第八条。Data export by an AI product is now tiered under these Provisions: below 100,000 people of non-sensitive personal information, the three procedures may be exempted; the band from 100,000 to below one million goes by standard contract or certification; the band at or above one million people, or at or above 10,000 people of sensitive personal information, or involving important data, requires a security assessment filing. The textual basis is Articles 5, 7 and 8.

R-18已生效 · 与 AI 的具体列管对应待核In force · specific AI listing correspondence pending

出口管制法Export Control Law

规则名Official name《中华人民共和国出口管制法》,中华人民共和国主席令第五十八号Export Control Law of the People's Republic of China, Order No. 58 of the President of the People's Republic of China

法域Jurisdiction中国China

方向Direction出境门Outbound gate

对应卡点EXP blocker监管审批(只记与 AI 技术出口直接相关口径)Regulatory approval (only the readings directly bearing on AI technology export are recorded)

已出台Promulgated

2020 年 10 月 17 日通过并公布。主席令载明已由第十三届全国人民代表大会常务委员会第二十二次会议于 2020 年 10 月 17 日通过,现予公布。Adopted and promulgated on October 17, 2020. The Presidential Order records that the Law was adopted at the 22nd session of the Standing Committee of the 13th National People's Congress on October 17, 2020, and is promulgated.

已生效In force

2020 年 12 月 1 日起施行,见主席令及第四十九条。In force from December 1, 2020, per the Presidential Order and Article 49.

执法已动Enforcement

有。商务部等部门依据该法及《中华人民共和国两用物项出口管制条例》发布 2024 年第 51 号公告,公布《中华人民共和国两用物项出口管制清单》自 2024 年 12 月 1 日起实施,许可管理已按清单运行。Yes. The Ministry of Commerce and other departments, acting under this Law and the Regulations on Export Control of Dual-Use Items, issued Announcement No. 51 of 2024, publishing the Dual-Use Items Export Control List of the People's Republic of China, implemented from December 1, 2024; licensing administration now runs by the list.

待核Pending

本法原文未点名人工智能。两用物项清单与 AI 的具体对应关系,官方答记者问只说明清单按 10 大行业领域、5 种物项类型系统整合现行已管制两用物项,数量约 700 项左右,未在该官方文本中核到以人工智能命名的条目。具体某项 AI 技术、软件或设备是否列管,须逐项查清单原文编码与参数,本条目标待核,不作推断。The Law's text does not name artificial intelligence. On the correspondence between the dual-use list and AI, the official press briefing states only that the list systematically consolidates currently controlled dual-use items across ten major industry fields and five item types, totalling roughly 700 items; no entry named for artificial intelligence was found in that official text. Whether a specific AI technology, software or equipment is listed must be checked item by item against the list's codes and parameters. This correspondence is marked pending, and no inference is drawn.

与 AI 技术出口直接相关的原文口径(仅限原文可支撑部分)Original-text readings directly bearing on AI technology export (only what the text can carry)

管制物项含技术,第二条定,国家对两用物项、军品、核以及其他与维护国家安全和利益、履行防扩散等国际义务相关的货物、技术、服务等物项的出口管制适用本法,管制物项包括物项相关的技术资料等数据。出口的定义含向外国组织和个人提供,第二条定,出口管制包括对从中华人民共和国境内向境外转移管制物项,以及中华人民共和国公民、法人和非法人组织向外国组织和个人提供管制物项,采取禁止或者限制性措施。两用物项定义,第二条定,既有民事用途,又有军事用途或者有助于提升军事潜力,特别是可以用于设计、开发、生产或者使用大规模杀伤性武器及其运载工具的货物、技术和服务。清单与许可制度,第四条、第九条、第十二条定,国家实行统一的出口管制制度,通过制定管制清单、实施出口许可等方式管理,国家出口管制管理部门制定与公布管制清单,清单内或临时管制物项出口须申请许可,清单外物项在出口经营者知道或应当知道可能危害国家安全和利益等法定风险时亦须申请许可。Controlled items include technology. Article 2 provides that this Law applies to state export control over dual-use items, military products, nuclear items, and other goods, technologies, services and items connected with safeguarding national security and interests and performing international obligations such as non-proliferation; controlled items include data such as technical materials related to the items. The definition of export includes provision to foreign organisations and individuals: Article 2 provides that export control covers prohibitive or restrictive measures on transferring controlled items from inside the People's Republic of China to outside it, and on citizens, legal persons and unincorporated organisations of the People's Republic of China providing controlled items to foreign organisations and individuals. Dual-use items are defined at Article 2 as goods, technologies and services that have both civilian uses and military uses, or that help enhance military potential, in particular items usable in designing, developing, producing or using weapons of mass destruction and their means of delivery. On lists and licensing, Articles 4, 9 and 12 provide that the state runs a unified export control system, administered through control lists and export licensing; the national export control authorities draw up and publish the control lists; export of items on a list, or under temporary control, requires a licence application; and for off-list items, a licence must also be applied for where the exporter knows, or ought to know, of statutory risks such as possible harm to national security and interests.

实际关口Practical gateAI 产品本身不因本法被整体管制,只有其货物、技术、服务或相关技术资料数据被列入管制清单、被实施临时管制,或落入第十二条清单外法定风险情形时,向境外转移或向外国组织和个人提供才进入出口许可关口。An AI product is not controlled as a whole by this Law. Only where its goods, technology, services or related technical-material data are placed on a control list, subjected to temporary control, or fall within the Article 12 off-list statutory risk cases does transfer abroad, or provision to foreign organisations and individuals, enter the export licensing gate.

三、入华向Part III · Inbound to China

R-19已生效 · 备案持续公布In force · filings published continuously

生成式人工智能服务管理暂行办法Interim Measures for the Management of Generative AI Services

规则名Official name《生成式人工智能服务管理暂行办法》,国家互联网信息办公室、国家发展和改革委员会、教育部、科学技术部、工业和信息化部、公安部、国家广播电视总局令第 15 号Interim Measures for the Management of Generative Artificial Intelligence Services, Order No. 15, issued jointly by the Cyberspace Administration of China, the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker监管审批 / 架构与部署Regulatory approval / Architecture and deployment

已出台Promulgated

令文称已经 2023 年 5 月 23 日国家网信办 2023 年第 12 次室务会会议审议通过并经其余六部门同意。公布日双口径分开记:令落款为 2023 年 7 月 10 日,网信办官网于 2023 年 7 月 13 日刊发全文与发布新闻,引用时须注明采用哪一口径。The order records deliberation and adoption at the CAC's 12th office meeting of 2023, on May 23, 2023, with the agreement of the other six departments. The two publication datings are kept apart: the order is dated July 10, 2023, and the CAC website published the full text and the release news on July 13, 2023; citations must state which dating is used.

已生效In force

原文第二十四条及令文均载,自 2023 年 8 月 15 日起施行。Article 24 of the text and the order both state effect from August 15, 2023.

执法已动Enforcement

有。网信部门按本办法持续开展生成式人工智能服务备案工作并发布已备案信息公告,官方公告称截至 2025 年 12 月 31 日累计 748 款生成式人工智能服务完成备案,435 款应用或功能完成登记。此处之备案为本办法项下的管理运行口径,其与 EXP-002 中生成式 AI 服务备案的对应关系以 EXP 条目内记述为准。Yes. The cyberspace authorities continuously carry out generative AI service filing work under these Measures and publish notices of filed services; an official notice states that, as of December 31, 2025, a cumulative 748 generative AI services had completed filing and 435 applications or functions had completed registration. The filing here is the administrative operation under these Measures; its correspondence to the generative AI service filing in EXP-002 follows the record inside the EXP entry.

待核Pending

无日期待核;公布日双口径已分开记。No date is pending; the two publication datings are recorded separately.

备案口径一句The filing rule, in one line

原文第十七条规定,提供具有舆论属性或者社会动员能力的生成式人工智能服务的,应当按照国家有关规定开展安全评估,并按照《互联网信息服务算法推荐管理规定》履行算法备案和变更、注销备案手续。Article 17 provides that a provider of a generative AI service with public-opinion attributes or social mobilisation capacity must carry out a security assessment in accordance with the relevant state provisions, and must complete algorithm filing, and change and cancellation filing procedures, in accordance with the Provisions on the Management of Algorithmic Recommendations in Internet Information Services.

实际关口Practical gate向中华人民共和国境内公众提供生成文本、图片、音频、视频等内容的服务即落入适用范围,未向境内公众提供服务的研发应用不适用本办法。A service providing generated text, images, audio, video and similar content to the public inside the People's Republic of China falls within the scope; research and development applications that do not provide services to the domestic public are outside these Measures.

R-20已生效 · 算法备案公示中In force · algorithm filings public

互联网信息服务算法推荐管理规定Provisions on the Management of Algorithmic Recommendations in Internet Information Services

规则名Official name《互联网信息服务算法推荐管理规定》,国家互联网信息办公室、工业和信息化部、公安部、国家市场监督管理总局令第 9 号Provisions on the Management of Algorithmic Recommendations in Internet Information Services, Order No. 9, issued jointly by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security and the State Administration for Market Regulation

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker监管审批Regulatory approval

已出台Promulgated

令文称已经 2021 年 11 月 16 日国家网信办 2021 年第 20 次室务会议审议通过并经其余三部门同意。公布日双口径分开记:令落款为 2021 年 12 月 31 日,网信办官网于 2022 年 1 月 4 日刊发全文与发布新闻。The order records deliberation and adoption at the CAC's 20th office meeting of 2021, on November 16, 2021, with the agreement of the other three departments. The two publication datings are kept apart: the order is dated December 31, 2021, and the CAC website published the full text and the release news on January 4, 2022.

已生效In force

原文第三十五条载,自 2022 年 3 月 1 日起施行。Article 35 of the text states effect from March 1, 2022.

执法已动Enforcement

有。原文第二十五条规定网信部门对材料齐全的备案予以备案、发放备案编号并进行公示,此后算法备案信息通过互联网信息服务算法备案系统持续公示,深度合成等专项备案批次亦援引本规定办理。Yes. Article 25 provides that the cyberspace authority files complete submissions, issues a filing number and makes the filing public; since then, algorithm filing information has been continuously published through the Internet Information Service Algorithm Filing System, and special filing batches such as deep synthesis are also handled under these Provisions.

待核Pending

无日期待核;公布日双口径已分开记。No date is pending; the two publication datings are recorded separately.

算法备案口径一句The algorithm filing rule, in one line

原文第二十四条规定,具有舆论属性或者社会动员能力的算法推荐服务提供者应当在提供服务之日起十个工作日内通过互联网信息服务算法备案系统填报备案信息、履行备案手续,变更与终止亦须按期办理变更、注销手续。Article 24 provides that an algorithm recommendation service provider with public-opinion attributes or social mobilisation capacity must, within ten working days of starting to provide the service, submit its filing information through the Internet Information Service Algorithm Filing System and complete the filing procedures; changes and terminations must likewise be filed for change or cancellation on schedule.

实际关口Practical gate在境内利用生成合成类、个性化推送类、排序精选类、检索过滤类、调度决策类等算法技术向用户提供信息,即须履行信息服务规范、用户权益保护与相应的备案和安全评估义务。Using algorithmic technologies, such as generative synthesis, personalised push, ranking and selection, retrieval and filtering, or scheduling and decision types, to provide information to users inside China triggers the duties on information service standards, user rights protection, and the corresponding filing and security assessment.

R-21已生效 · 备案批次持续公布In force · filing batches published continuously

互联网信息服务深度合成管理规定Provisions on the Management of Deep Synthesis in Internet Information Services

规则名Official name《互联网信息服务深度合成管理规定》,国家互联网信息办公室、工业和信息化部、公安部令第 12 号Provisions on the Management of Deep Synthesis in Internet Information Services, Order No. 12, issued jointly by the Cyberspace Administration of China, the Ministry of Industry and Information Technology and the Ministry of Public Security

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker架构与部署 / 分发渠道Architecture and deployment / Distribution channels

已出台Promulgated

令文称已经 2022 年 11 月 3 日国家网信办 2022 年第 21 次室务会议审议通过并经工业和信息化部、公安部同意。公布日双口径分开记:令落款为 2022 年 11 月 25 日,网信办官网于 2022 年 12 月 11 日刊发全文与发布新闻。The order records deliberation and adoption at the CAC's 21st office meeting of 2022, on November 3, 2022, with the agreement of the Ministry of Industry and Information Technology and the Ministry of Public Security. The two publication datings are kept apart: the order is dated November 25, 2022, and the CAC website published the full text and the release news on December 11, 2022.

已生效In force

原文第二十五条载,自 2023 年 1 月 10 日起施行。Article 25 of the text states effect from January 10, 2023.

执法已动Enforcement

有。国家网信办已按本规定持续公开发布境内深度合成服务算法备案信息,官网可见第十八批、第十九批等批次公告,并提示未备案者尽快申请备案。Yes. The CAC has continuously published domestic deep synthesis service algorithm filing information under these Provisions; its website carries batch notices such as the eighteenth and nineteenth batches, and reminds those not yet filed to apply for filing as soon as possible.

待核Pending

无日期待核;公布日双口径已分开记。No date is pending; the two publication datings are recorded separately.

标识口径一句The marking rule, in one line

原文第十六条要求对生成或者编辑的信息内容采取技术措施添加不影响用户使用的标识,第十七条要求智能对话、合成人声、人脸生成、沉浸式拟真场景等可能导致公众混淆或者误认的服务在合理位置、区域进行显著标识,第十八条禁止删除、篡改、隐匿标识。Article 16 requires technical measures to add marks to generated or edited information content that do not affect user use; Article 17 requires services that may cause public confusion or misrecognition, such as intelligent dialogue, synthetic voice, face generation and immersive realistic scenes, to carry conspicuous marks in reasonable positions and areas; Article 18 prohibits deleting, tampering with or concealing the marks.

实际关口Practical gate在境内应用深度合成技术提供互联网信息服务即适用本规定,且应用程序分发平台须核验深度合成类应用程序的安全评估、备案等情况,违规可被不予上架、暂停服务或者下架。Applying deep synthesis technology to provide internet information services inside China falls under these Provisions, and application distribution platforms must verify the security assessment and filing status of deep synthesis applications; violations can lead to refusal of listing, suspension of service or takedown.

R-22已生效(2025-09-01)· 执法通报已出In force (Sep 1, 2025) · enforcement notice issued

人工智能生成合成内容标识办法Measures for Labeling AI-Generated and Synthesized Content

规则名Official name《人工智能生成合成内容标识办法》,国家互联网信息办公室、工业和信息化部、公安部、国家广播电视总局制定,印发通知文号国信办通字〔2025〕2 号;配套强制性国家标准 GB 45438-2025《网络安全技术 人工智能生成合成内容标识方法》(英文名 Cybersecurity technology, Labeling method for content generated by artificial intelligence)Measures for Labeling Content Generated and Synthesized by Artificial Intelligence, issued by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration, under issuance notice Guo Xin Ban Tong Zi [2025] No. 2; with the accompanying mandatory national standard GB 45438-2025, Cybersecurity technology, Labeling method for content generated by artificial intelligence

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker架构与部署 / 分发渠道Architecture and deployment / Distribution channels

已出台Promulgated

印发日与发布日分开记:通知落款为 2025 年 3 月 7 日,网信办官网于 2025 年 3 月 14 日刊发通知全文,同日刊发四部门联合发布新闻。配套国标经国家标准平台载,发布日期 2025 年 2 月 28 日。The issuance date and the publication date are recorded separately: the notice is dated March 7, 2025, and the CAC website published the full notice on March 14, 2025, publishing the four departments' joint release news the same day. The accompanying national standard, as recorded on the national standards platform, was published on February 28, 2025.

已生效In force

原文第十四条载,自 2025 年 9 月 1 日起施行;配套国标实施日期同为 2025 年 9 月 1 日,与本办法同步实施。Article 14 of the text states effect from September 1, 2025; the accompanying national standard's implementation date is likewise September 1, 2025, implemented in step with these Measures.

执法已动Enforcement

有。网信办官网 2025 年 11 月 25 日通报,网信部门依法集中查处一批存在人工智能生成合成内容标识违法违规问题的移动互联网应用程序,处置包括约谈、责令限期改正、下架下线。Yes. A CAC notice of November 25, 2025 records that the cyberspace authorities, acting in accordance with law, dealt in a concentrated action with a batch of mobile internet applications with labeling violations for AI-generated and synthesized content; the measures taken included regulatory interviews, orders to rectify within a time limit, and takedown and removal.

待核Pending

无日期待核;印发日与发布日已分开记。No date is pending; the issuance and publication dates are recorded separately.

标识口径一句The labeling rule, in one line

标识分显式标识与隐式标识,服务提供者须在生成合成内容中添加用户可明显感知的显式标识,并在文件元数据中添加含内容属性、服务提供者名称或者编码、内容编号等要素的隐式标识,任何组织和个人不得恶意删除、篡改、伪造、隐匿标识。Labels are divided into explicit and implicit labels: a service provider must add an explicit label, clearly perceptible to users, to generated and synthesized content, and must add an implicit label in the file metadata carrying elements such as the content attributes, the service provider's name or code and the content number; no organisation or individual may maliciously delete, tamper with, forge or conceal a label.

实际关口Practical gate生成合成服务、内容传播服务与应用程序分发平台被连成同一标识链,分发平台在上架或者上线审核时须要求服务提供者说明是否提供人工智能生成合成服务并核验其标识相关材料。Generation and synthesis services, content distribution services and application distribution platforms are linked into a single labeling chain: at listing or launch review, a distribution platform must require the service provider to state whether it provides AI generation and synthesis services, and must verify its labeling materials.

R-23已生效 · 2025 年修正决定 2026-01-01 施行In force · the 2025 amendment decision takes effect Jan 1, 2026

网络安全法(含 2025 年修正)Cybersecurity Law (including the 2025 amendment)

规则名Official name《中华人民共和国网络安全法》Cybersecurity Law of the People's Republic of China

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker监管审批 / 数据与隐私合规Regulatory approval / Data and privacy compliance

已出台Promulgated

原法 2016 年 11 月 7 日第十二届全国人民代表大会常务委员会第二十四次会议通过(2025 年修正版全文页眉载)。The original Law was adopted on November 7, 2016 at the 24th session of the Standing Committee of the 12th National People's Congress (as carried in the header of the full text of the 2025 amended version).

已生效In force

原法自 2017 年 6 月 1 日起施行(修正版全文第八十一条仍载)。2025 年修正单独记:第十四届全国人民代表大会常务委员会第十八次会议于 2025 年 10 月 28 日通过《关于修改〈中华人民共和国网络安全法〉的决定》,决定原文末句载本决定自 2026 年 1 月 1 日起施行,法律根据本决定作相应修改并对条文顺序作相应调整、重新公布。截至本条核查日修正决定已生效,与原法 2017 年施行日不冲突,两者分开记。修正新增第二十条涉及人工智能,原文为国家支持人工智能基础理论研究和算法等关键技术研发、完善伦理规范、加强风险监测评估和安全监管。The original Law has been in force since June 1, 2017 (Article 81 of the amended text still so states). The 2025 amendment is recorded separately: on October 28, 2025, the 18th session of the Standing Committee of the 14th National People's Congress adopted the Decision on Amending the Cybersecurity Law of the People's Republic of China; the Decision's final sentence states that it takes effect on January 1, 2026, and that the Law is correspondingly amended, its article order correspondingly adjusted, and the Law republished. As of this entry's check date the amendment Decision is in effect; it does not conflict with the original Law's 2017 implementation date, and the two are recorded separately. The amendment's new Article 20 concerns artificial intelligence, providing in the original text that the state supports basic theoretical research on AI and research and development of key technologies such as algorithms, improves ethical norms, and strengthens risk monitoring, assessment and security regulation.

执法已动Enforcement

有制度性执法安排。修正版全文设法律责任专章并提高部分罚款幅度,网信、电信、公安等部门依职责负责监督管理。具体单项执法个案通报未在本轮单独核,列入待核清单。There is an institutional enforcement arrangement: the amended text carries a dedicated legal-liability chapter and raises some fine levels, and the cyberspace, telecommunications, public security and other departments supervise and administer within their respective remits. Individual enforcement case notices were not separately checked this round and are placed on the pending list.

待核Pending

无日期待核;仅具体执法个案未另核,不影响通过与施行日期的认定。No date is pending; only the individual enforcement cases were not separately checked, which does not affect the findings on adoption and implementation dates.

注Note

本轮曾尝试打开中国人大网修改决定另一原页,页面抓取失败,未以该页为据;上列人大网专题页已完整载明通过日期与决定施行日期,故修正事项不标待核。This round also tried to open another original National People's Congress page for the amendment Decision; the fetch failed and that page was not relied on. The NPC topic page listed above carries the adoption date and the Decision's effective date in full, so the amendment is not marked pending.

实际关口Practical gate在境内建设、运营、维护和使用网络以及网络安全的监督管理适用本法,网络运营者须履行等级保护、真实身份、日志留存等义务;关键信息基础设施运营者在境内收集和产生的个人信息和重要数据应当在境内存储,确需向境外提供须按办法进行安全评估。The construction, operation, maintenance and use of networks inside China, and the supervision and administration of cybersecurity, fall under this Law; network operators must meet duties such as classified protection, real-identity requirements and log retention. Personal information and important data that critical information infrastructure operators collect and generate inside China must be stored inside China; where provision abroad is genuinely needed, a security assessment must follow the applicable measures.

R-24已生效In force

数据安全法Data Security Law

规则名Official name《中华人民共和国数据安全法》Data Security Law of the People's Republic of China

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker数据与隐私合规 / 监管审批Data and privacy compliance / Regulatory approval

已出台Promulgated

2021 年 6 月 10 日第十三届全国人民代表大会常务委员会第二十九次会议通过,现予公布(中国人大网全文页与主席令第八十四号载)。Adopted on June 10, 2021 at the 29th session of the Standing Committee of the 13th National People's Congress, and promulgated (as carried on the NPC website's full-text page and in Presidential Order No. 84).

已生效In force

原文第五十五条载,本法自 2021 年 9 月 1 日起施行。Article 55 of the text states that this Law takes effect on September 1, 2021.

执法已动Enforcement

有制度性执法安排。原文设数据安全审查、风险评估、约谈整改与法律责任,重要数据出境与向境外提供重要数据设有专门罚则。具体单项执法个案通报未在本轮单独核,列入待核清单。There is an institutional enforcement arrangement: the text sets up data security review, risk assessment, regulatory interviews and rectification, and legal liability, with dedicated penalties for important-data export and for providing important data abroad. Individual enforcement case notices were not separately checked this round and are placed on the pending list.

待核Pending

通过、公布、施行日期无待核;仅具体执法个案未另核。The adoption, promulgation and implementation dates carry nothing pending; only the individual enforcement cases were not separately checked.

实际关口Practical gate在境内开展数据处理活动适用本法,训练与运营数据须按分类分级保护;重要数据处理者须定期开展风险评估;重要数据出境与向外国司法或者执法机构提供境内数据均受专门管制(境外提供境内数据的审批口径以原文第二十一条、第三十六条为准,本条只记关口不展开条文复述)。Data processing activities carried out inside China fall under this Law: training and operational data must be protected by category and grade; important-data processors must periodically carry out risk assessments; and the export of important data, and the provision of domestically held data to foreign judicial or law-enforcement bodies, are specially controlled (the approval reading for providing domestic data abroad follows Articles 21 and 36 of the text; this entry records the gate only and does not expand the provisions).

R-25已生效 · 有域外适用条款In force · with extraterritorial application

个人信息保护法Personal Information Protection Law

规则名Official name《中华人民共和国个人信息保护法》Personal Information Protection Law of the People's Republic of China

法域Jurisdiction中国China

方向Direction入华Inbound

对应卡点EXP blocker数据与隐私合规 / 监管审批Data and privacy compliance / Regulatory approval

已出台Promulgated

2021 年 8 月 20 日第十三届全国人民代表大会常务委员会第三十次会议通过(中国人大网全文载)。Adopted on August 20, 2021 at the 30th session of the Standing Committee of the 13th National People's Congress (as carried in the NPC website's full text).

已生效In force

原文第七十四条载,本法自 2021 年 11 月 1 日起施行。Article 74 of the text states that this Law takes effect on November 1, 2021.

执法已动Enforcement

有制度性执法安排。原文设履行个人信息保护职责的部门、合规审计、个人信息保护影响评估与法律责任专章,情节严重可处五千万元以下或者上一年度营业额百分之五以下罚款。具体单项执法个案通报未在本轮单独核,列入待核清单。There is an institutional enforcement arrangement: the text establishes the departments that perform personal information protection duties, compliance audits, personal information protection impact assessments and a dedicated legal-liability chapter, under which serious cases can be fined up to RMB 50 million or up to 5% of the previous year's turnover. Individual enforcement case notices were not separately checked this round and are placed on the pending list.

待核Pending

通过、施行日期无待核;仅具体执法个案未另核。The adoption and implementation dates carry nothing pending; only the individual enforcement cases were not separately checked.

实际关口Practical gate境外处理者以向境内自然人提供产品或者服务为目的、或者分析评估境内自然人行为的,亦适用本法,并须在境内设立专门机构或者指定代表;个人信息跨境提供须满足安全评估、保护认证、标准合同等条件之一并取得单独同意(域外适用与境内代表要求见原文第三条、第五十三条,本条只记关口)。An overseas processor that processes for the purpose of providing products or services to natural persons inside China, or of analysing and assessing the behaviour of natural persons inside China, is likewise covered by this Law, and must establish a dedicated body inside China or designate a representative. Cross-border provision of personal information must satisfy one of the conditions, security assessment, protection certification or standard contract, and must obtain separate consent (the extraterritorial application and the domestic-representative requirement are at Articles 3 and 53 of the text; this entry records the gate only).

官方来源Official sources

核查日期 2026-11-01Checked November 1, 2026

四、待核项清单(缺口与原因)Part IV · Pending-verification annex (items and the reasons for their gaps)

条目Item待核内容What is pending缺口原因Why the gap
R-042026-06-30 至 2027-01-01 间原条文可执行状态;涉原法的联邦诉讼对执行的影响The enforceability of the original provisions between June 30, 2026 and January 1, 2027; the effect of federal litigation over the original Act on its enforcement未打开官方过渡条款与法院文件核实The official transitional provisions and court documents were not opened and checked
R-05是否已有首例执法个案Whether a first enforcement case exists本次打开的官方页面未见已公布处罚个案The official pages opened this round showed no published penalty case
R-06具体执法个案清单与罚款额The list of individual enforcement cases and their fine amounts未逐个打开 CPPA 个案页核The CPPA's case pages were not opened and checked one by one
R-07执法路径是否经不正当竞争法执行;针对该法的诉讼进展Whether the enforcement path runs through unfair competition law; litigation progress against the Act原文未设罚则与执法机关条款,无官方原文支撑The text sets no penalty or enforcement-authority provision, and no official text supports a reading
R-09第五条原始制定年份(FTC 官网史页口径)The original year of enactment of Section 5 (on the FTC's official history-page reading)本次未打开 FTC 官网史页核The FTC's official history page was not opened this round
R-102027 年施行的一批 PIPA 修订条款与现行条文对应;PIPC 处罚个案How the batch of PIPA amendments taking effect in 2027 maps onto the current provisions; PIPC penalty cases版本滚动快,个案页未逐个打开The text rolls forward quickly, and the case pages were not opened one by one
R-112026 年 1 月 20 日部分修订(法律第 21311 号)新增条款分批施行日期;处罚个案有无The staged effective dates of the provisions added by the January 20, 2026 partial amendment (Act No. 21311); whether any penalty case exists未逐条对 law.go.kr 附则核完The law.go.kr Addenda were not checked through article by article
R-12APPI 引用条文所据版本日期;PPC 命令或劝告个案The version date relied on for APPI citations; PPC order or recommendation cases滚动修订中,个案页未逐个打开Under rolling amendment, and the case pages were not opened one by one
R-13LGPD 其余条文实际操作生效日(2020-09-18 一说)The operative effective date of the LGPD's remaining provisions (one account says September 18, 2020)仅凭汇编本 Article 65 读不干净,须以 DOU 最终文本逐字复核;Planalto 原页直开遭拦截It cannot be read cleanly from Article 65 of the consolidated text alone and must be re-verified word by word against the final DOU text; the Planalto original page was blocked on direct opening
R-14DPDP 一年段与十八个月段的具体日历日;法案附表罚金上限The exact calendar dates of the DPDP one-year and eighteen-month tranches; the penalty ceilings in the Act's ScheduleG. 843(E) 只写相对期间且公报刊期日与文件号日期有一日之差;罚金上限未在法案原文页逐项核到G. 843(E) states only relative periods, and the gazette issue date differs by one day from the date in the document number; the penalty ceilings were not verified item by item on the Act's text page
R-16公布日口径(令文落款 2023-02-22 与官网刊载 2023-02-24 并存)The publication dating (the order's date of February 22, 2023 and the website posting of February 24, 2023 coexist)两处官方文本表述不同,已分开记,引用须注明口径Two official texts state it differently; the two are recorded separately and citations must state the dating used
R-17各自贸试验区负面清单的具体内容The specific contents of each pilot free trade zone's negative list清单为各区分别制定与备案,不在本规定原文内The lists are drawn up and filed zone by zone, and are not in the text of these Provisions
R-18具体 AI 技术、软件或设备是否列入两用物项管制清单Whether specific AI technologies, software or equipment are on the dual-use items control list本法与官方答记者问未点名人工智能,须逐项查清单原文编码与参数Neither the Law nor the official press briefing names artificial intelligence; the list's codes and parameters must be checked item by item
R-23 至 R-25三法各自的具体单项执法个案通报The individual enforcement case notices under each of the three laws本轮只核到制度性执法安排,个案未单独核This round verified only the institutional enforcement arrangements; the individual cases were not separately checked

编制说明Compilation notes

  1. 任务单所列联邦与州级美国条目中,加州 AI 透明类按 AB 2013 与 SB 942 两部法分列;得州 TRAIGA 的成法载体经官方进程核为 HB 149,HB 1709 未成法,已在条目内纠正。Among the federal and state-level US items, the California AI transparency line is split into two statutes, AB 2013 and SB 942. The enacted vehicle of Texas's TRAIGA was verified against the official record as HB 149; HB 1709 was never enacted, and R-05 corrects the point inside the entry.
  2. 状态纪律:出台、生效、执法已动、待核四格分开。本轮查出三处「通行说法与现行官方文本不符」并已在条目内改正单写口径:欧盟 AI 法案高风险适用日已被 Regulation (EU) 2026/1744 推迟;科罗拉多 AI 法义务日经 SB 25B-004 推迟、又经 SB 26-189 重订;加州 SB 942 施行日经 AB 853 推迟至 2026 年 8 月 2 日。Status discipline: promulgated, in force, enforcement and pending are four separate cells. This round found three points where common accounts conflict with the current official texts, and corrected the single-date readings inside the entries: the EU AI Act's high-risk application dates were postponed by Regulation (EU) 2026/1744; the Colorado AI Act's duty dates were delayed by SB 25B-004 and then re-made by SB 26-189; and SB 942's operative date was delayed by AB 853 to August 2, 2026.
  3. 执法案例只入条目「执法已动」格与 R-02 的执法注脚,不改写规则本体;Garante 对 DeepSeek 的正式决定只作 GDPR 的执法注脚出现,与 EXP-003 条目互指,不在本清单内另立条目。Enforcement cases sit only in an entry's enforcement cell and in R-02's enforcement footnote; they do not rewrite the rule itself. The Garante's formal decision on DeepSeek appears only as the GDPR's enforcement footnote, cross-referring to entry EXP-003, and is not given its own entry in this checklist.
  4. 下一轮补核优先级:R-14 的两段日历日与罚金附表、R-13 的 DOU 最终文本、R-04 的过渡条款原文。Next round's verification priorities: R-14's two tranche calendar dates and its penalty Schedule; R-13's final DOU text; and R-04's transitional provisions in the original.

本页为《AI 产品出海与入华规则清单》v0.1(2026-11-01 首编)站上呈现,核查截至 2026-11-01,归属一号台(企业 AI 架构情报台),为《AI 产品出海登记册》附件,署名深翼 DeepWing。规则状态变化走修订记录,不覆盖原文。This page presents the Rules Checklist for AI Products Going Global and Entering China, v0.1 (first edition, November 1, 2026), checked as of November 1, 2026, at Desk 1 (Enterprise AI Architecture), as an annex to the AI Global Expansion Register, by DeepWing. Changes to a rule's status are kept as revision records; the original text is not overwritten.