US, China and EU pursue distinct paths to AI regulation — Eter News

Enriched Analysis 深度解析

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

The US is relying mainly on voluntary industry pledges and executive orders, while Congress has not passed a federal AI law. China combines evolving rules, content controls and calls for human oversight without endorsing a broad slowdown. The EU has enacted a comprehensive AI Act but delayed some high-risk rules amid concerns about innovation and competitiveness.美国主要依靠行业自愿承诺和行政命令监管人工智能，但国会尚未通过联邦人工智能法律。中国结合不断演变的监管规定、内容管控和“始终由人控制”的要求，同时不支持全面放缓发展。欧盟已通过综合性的《人工智能法案》，但在创新和竞争力担忧下推迟了部分高风险系统规则。

Taipei Times 7 Oct 2026 2026年10月7日

9090

US AI Action Plan measures美国人工智能行动计划措施

3030

Voluntary early-access period for powerful models强大模型自愿提前访问期限

01The Story新闻事实

The US is relying mainly on voluntary industry pledges and executive orders, while Congress has not passed a federal AI law. China combines evolving rules, content controls and calls for human oversight without endorsing a broad slowdown. The EU has enacted a comprehensive AI Act but delayed some high-risk rules amid concerns about innovation and competitiveness.美国主要依靠行业自愿承诺和行政命令监管人工智能，但国会尚未通过联邦人工智能法律。中国结合不断演变的监管规定、内容管控和“始终由人控制”的要求，同时不支持全面放缓发展。欧盟已通过综合性的《人工智能法案》，但在创新和竞争力担忧下推迟了部分高风险系统规则。

  • The US has not passed a federal AI law, while state AI laws are taking effect nationwide.美国尚未通过联邦人工智能法律，而各州的人工智能法律正在全国范围内生效。
  • The White House’s AI Action Plan listed more than 90 measures to accelerate development.白宫的人工智能行动计划列出了90多项加快发展的措施。
  • China’s rules on AI companion bots took effect in July, including restrictions on emotional dependency and unauthorized digital clones.中国关于人工智能陪伴机器人的规定于7月生效，包括限制情感依赖，以及禁止未经同意制作真人数字克隆体。
  • Chinese leaders have called for human control, stronger laws, technological monitoring, early warning and emergency response systems.中国领导人呼吁加强人工控制、法律法规、技术监测、预警和应急响应体系。
  • 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》，之后推迟了涉及高风险系统的规定。

The three approaches differ over how to balance AI development, safety, content control and protection of vulnerable users.三种路径在如何平衡人工智能发展、安全、内容管控以及对弱势用户的保护方面存在差异。

023 Dimensions维度分析

Comparison对比

Regulatory architecture监管架构

The US is using executive action and voluntary commitments, China is applying an evolving set of targeted rules and controls, and the EU is relying on a formal legal framework.美国依靠行政行动和自愿承诺，中国采用不断演变的针对性规定与管控措施，欧盟则依靠正式的法律框架。

US AI Action Plan measures美国人工智能行动计划措施

July last year去年7月90 measures

EU AI Act adoption year欧盟《人工智能法案》通过年份2,024 year

  • The US has not enacted a federal AI law, while the EU adopted its AI Act in 2024.美国尚未制定联邦人工智能法律，而欧盟在2024年通过了《人工智能法案》。
  • China has a complex, evolving set of rules governing AI use, including by children.中国已经建立了一套复杂且不断演变的人工智能使用规则，其中包括涉及儿童的规定。
Comparison对比

Safety and control安全与控制

All three systems address safety, but through different mechanisms: US developers may provide early access voluntarily, China emphasizes human control and security, and the EU imposes legal scrutiny on selected services.三方都处理人工智能安全问题，但机制不同：美国开发商可自愿提供模型早期访问，中国强调人工控制和国家安全，欧盟则对部分服务实施法律审查。

Voluntary early-access period for powerful models强大模型自愿提前访问期限

This year今年30 days

  • A June US order created a voluntary system allowing developers to give the government up to 30 days of early access to their most powerful models.美国6月发布的一项行政命令建立了自愿机制，允许开发商向政府提供其最强大模型最多30天的提前访问权限。
  • China’s president said AI should be “always under human control.”中国国家主席表示，人工智能应当“始终处于人的控制之下”。
  • ChatGPT became the first AI chatbot added to the EU’s list of digital services subject to greater legal scrutiny.ChatGPT成为首个被列入欧盟数字服务名单、并接受更严格法律审查的人工智能聊天机器人。
Stakeholder impact利益相关方

Children and emotional dependency儿童与情感依赖

China and the EU are both moving to address risks from AI companions, including emotional dependency and unauthorized digital replicas. The EU’s proposal would require safeguards before children use such systems.中国和欧盟都在应对人工智能陪伴工具带来的风险，包括情感依赖和未经授权的数字复制品。欧盟提议要求在儿童使用此类系统前建立安全措施。

  • China’s July rules ban creating digital clones of real people without their consent.中国7月实施的规定禁止未经本人同意制作真人数字克隆体。
  • The EU said AI chatbots must not simulate interpersonal relationships in ways that create emotional dependency.欧盟表示，人工智能聊天机器人不得以造成情感依赖的方式模拟人际关系。

03Timeline时间线

  1. 2024 The European Union adopted its AI Act.欧盟通过《人工智能法案》。

Key takeaway核心结论

AI governance is fragmenting into three models that place different weight on innovation, enforceable rules, security and user protection.人工智能治理正在分化为三种模式，分别对创新、可执行规则、安全和用户保护赋予不同权重。

04What to Watch后续关注

  1. Whether the US Congress passes a federal AI law after the midterm elections.美国国会是否会在中期选举后通过联邦人工智能法律。
  2. How US companies implement independent audits and board-level oversight under the White House accord.美国企业将如何落实白宫协议中的独立审计和董事会层级监督。
  3. Whether China expands rules for AI companions and other AI-generated content.中国是否会扩大对人工智能陪伴工具及其他人工智能生成内容的监管规定。
  4. How the EU’s delayed high-risk rules and proposed child safeguards affect companies and users.欧盟推迟的高风险系统规则及拟议的儿童保护措施将如何影响企业和用户。

05Uncertainties不确定因素

  • The US voluntary pact carries no penalties for companies that fall short.美国的自愿协议不包含对未达标企业的处罚。
  • The source does not specify when the EU’s delayed high-risk rules will take effect.来源没有说明欧盟推迟的高风险系统规则何时生效。
  • China’s policy direction supports regulation and control but does not endorse a general pause or deliberate slowdown in AI development.中国的政策方向支持监管和管控，但不支持全面暂停或有意放缓人工智能发展。

06Sources信息来源

  1. US, China, EU: three paths to regulating AI - Taipei TimesUS, China, EU: three paths to regulating AI - Taipei Times Taipei Times · 2026-10-07

Share this analysis 分享此深度分析

X Facebook LinkedIn WhatsApp