US, China and EU take different paths to regulate AI — Eter News

Enriched Analysis 深度解析

US, China and EU take different paths to regulate AI美国、中国与欧盟采取不同的人工智能监管路径

The United States is relying largely on voluntary industry pledges and executive orders, while China is combining detailed rules, content controls and calls for human oversight without endorsing a broad slowdown. The European Union has adopted comprehensive legislation but delayed some high-risk requirements as companies warn that regulation could weaken their competitiveness.美国主要依靠行业自愿承诺和行政命令监管人工智能；中国则结合具体规则、内容管控和对人工监督的要求，但并未支持广泛放缓发展。欧盟已通过较为全面的立法，但在企业担忧监管削弱竞争力后，推迟了部分高风险系统的要求。

AFP 7 Oct 2026 2026年10月7日

9090

Measures in the 2025 AI Action Plan2025年人工智能行动计划中的措施数量

01The Story新闻事实

The United States is relying largely on voluntary industry pledges and executive orders, while China is combining detailed rules, content controls and calls for human oversight without endorsing a broad slowdown. The European Union has adopted comprehensive legislation but delayed some high-risk requirements as companies warn that regulation could weaken their competitiveness.美国主要依靠行业自愿承诺和行政命令监管人工智能；中国则结合具体规则、内容管控和对人工监督的要求，但并未支持广泛放缓发展。欧盟已通过较为全面的立法，但在企业担忧监管削弱竞争力后，推迟了部分高风险系统的要求。

  • The United States has not yet enacted a federal AI law, while state-level AI laws are taking effect.美国尚未通过联邦人工智能法律，但各州的人工智能法律正在生效。
  • A July 2025 US AI Action Plan listed more than 90 measures intended to accelerate development.美国2025年7月发布的人工智能行动计划列出了90多项旨在加快发展的措施。
  • China requires AI companion rules to address emotional dependency and bans digital clones of real people without consent.中国针对人工智能陪伴产品的规则旨在遏制情感依赖，并禁止未经同意制作真实人物的数字克隆体。
  • China's policy posture supports regulation and human control but does not place a general pause or deliberate slowdown at the center of its approach.中国的政策立场支持监管和由人控制人工智能，但没有把全面暂停或有意放缓发展作为核心方针。
  • The EU adopted its AI Act in 2024, later delaying rules covering high-risk systems amid concerns about innovation and competitiveness.欧盟于2024年通过《人工智能法案》，随后因创新和竞争力方面的担忧推迟了涉及高风险系统的规则。

The three approaches differ mainly in how they balance innovation, state control, public safety and the enforceability of rules.三种路径的主要差异，在于如何平衡创新、国家管控、公共安全以及监管规则的可执行性。

023 Dimensions维度分析

Comparison对比

US: voluntary commitments and fragmented enforcement美国：自愿承诺与分散执法

Washington is prioritizing rapid AI development through executive action and voluntary industry agreements. The absence of a federal law leaves state rules in force and means recent commitments carry no stated penalties for non-compliance.华盛顿通过行政行动和行业自愿协议优先推动人工智能快速发展。由于联邦层面尚无法律，各州规则继续生效，而近期行业承诺对不遵守行为没有规定处罚。

Measures in the 2025 AI Action Plan2025年人工智能行动计划中的措施数量

90 more than measures90项以上

20252025年

Maximum early access offered to government向政府提供的最长提前访问时间

30 days30天

20262026年

  • Congress has yet to pass a federal AI law and has repeatedly refused to override state AI laws.国会尚未通过联邦人工智能法律，并多次拒绝推翻州级人工智能法律。
  • Major AI companies signed a September 29 accord pledging independent audits and board-level oversight.主要人工智能公司于9月29日签署协议，承诺开展独立审计并接受董事会层面的监督。
Contradiction矛盾之处

China: control without a broad development pause中国：加强管控但不全面放缓发展

China is expanding rules for specific uses and risks, including children, AI-generated content and national security. At the same time, analysts cited in the article say a general pause or deliberate slowdown remains outside the mainstream policy posture.中国正在针对儿童使用、人工智能生成内容和国家安全等具体用途与风险扩大监管。与此同时，文章援引的分析人士表示，全面暂停或有意放缓发展仍不属于中国主流政策立场。

  • Xi Jinping called for AI to remain under human control and urged stronger laws, monitoring, warnings and emergency response systems.习近平要求人工智能始终处于人的控制之下，并呼吁加强法律法规、技术监测、预警和应急响应体系。
  • China censors and controls AI-generated content in a manner similar to other online information.中国以类似其他网络信息的方式，对人工智能生成内容进行审查和管控。
Consequence影响推演

EU: comprehensive rules tempered by competitiveness concerns欧盟：全面规则受到竞争力担忧的制约

The EU presents its AI Act as a comprehensive framework and is adding scrutiny through online-content rules. However, concerns from European companies about innovation and competition have contributed to delays for some high-risk requirements.欧盟将《人工智能法案》定位为全面监管框架，并通过网络内容规则加强审查。不过，欧洲企业对创新和竞争的担忧促成了部分高风险要求的推迟。

  • The EU adopted its AI Act in 2024 and later delayed rules covering high-risk systems.欧盟于2024年通过《人工智能法案》，随后推迟了涉及高风险系统的规则。
  • ChatGPT became the first AI chatbot added to the EU's list of digital services subject to greater legal scrutiny in August.8月，ChatGPT成为首个人工智能聊天机器人，被列入欧盟需要接受更严格法律审查的数字服务名单。

03Timeline时间线

  1. 2023 The Biden administration issued an order requiring developers of the most powerful AI systems to share safety test results with the government.拜登政府发布行政命令，要求最强大人工智能系统的开发者向政府共享安全测试结果。
  2. 2024 The European Union adopted its AI Act.欧盟通过《人工智能法案》。
  3. 2025-07 The White House unveiled an AI Action Plan listing more than 90 measures.白宫公布人工智能行动计划，列出90多项措施。
  4. 2026-09-29 Major AI companies signed a White House accord pledging independent audits and board-level oversight.主要人工智能公司签署白宫协议，承诺开展独立审计并接受董事会层面的监督。
  5. 2026-10-07 The comparison of US, Chinese and EU AI regulation was published.有关美国、中国和欧盟人工智能监管路径的比较文章发布。

Key takeaway核心结论

The US, China and EU are regulating AI through three distinct mixes of market support, state control and legally binding oversight.美国、中国和欧盟正分别通过市场支持、国家管控和具有法律约束力的监督，形成三种不同的人工智能监管组合。

04What to Watch后续关注

  1. Whether the US Congress passes a federal AI law after the midterm elections.美国国会在中期选举后是否会通过联邦人工智能法律。
  2. Whether voluntary US company commitments lead to stronger oversight despite having no stated penalties.美国企业的自愿承诺在没有明确处罚的情况下，是否会带来更严格的监督。
  3. How China expands rules for AI-generated content, children and national security while maintaining development momentum.中国如何在保持发展势头的同时，扩大对人工智能生成内容、儿童使用和国家安全的监管。
  4. Whether the EU's delayed high-risk requirements and proposed child-safety measures are implemented.欧盟推迟的高风险系统要求和拟议的儿童安全措施是否会落地。

05Uncertainties不确定因素

  • The article does not specify when the EU's delayed high-risk requirements will take effect.文章没有说明欧盟推迟的高风险系统要求何时生效。
  • The White House accord describes company pledges but does not establish penalties for failing to meet them.白宫协议规定了企业承诺，但没有为未能履行承诺设立处罚。
  • The article does not quantify how many state AI laws are currently in effect in the United States.文章没有量化目前美国有多少州级人工智能法律正在生效。

06Sources信息来源

  1. US, China, EU: Three paths to regulating AIUS, China, EU: Three paths to regulating AI AFP · 2026-10-07

Share this analysis 分享此深度分析

X Facebook LinkedIn WhatsApp