2026年6月25日 下午2:03 UTC / 路透社
记者：卡伦·弗赖菲尔德
2026年6月25日 下午2:03 UTC 更新于19分钟前
2022年11月13日，美国华盛顿，在2022年中期选举后，当选为美国第118届国会议员的内森尼尔·莫兰（共和党-德克萨斯州）在凯悦酒店外出席新一届议员入职培训。路透社/迈克尔·A·麦科伊
内容提要
- 法案将要求向美国商务部通报相关情况
- 随着人工智能模型能力愈发强大，对人工智能监管的呼声日益高涨
- 针对性法案有望打破国会在人工智能立法方面的停滞僵局
6月25日（路透社）——一名共和党议员周四提出法案，要求人工智能模型开发者上报危险能力、安全漏洞和安全事件。
这项由美国德克萨斯州众议员内森尼尔·莫兰提出的草案法案，将强制要求人工智能企业在发现危险活动后的7天内向美国商务部通报，同时要求商务部在接到最严重事件报告后的48小时内通报国会。
开启您的晨间法律资讯订阅
接收《每日案卷》新闻简报的最新法律新闻，点击此处注册。
“这是一项早发现、拉警报的法案，”莫兰在谈及《人工智能事件上报法案》时接受采访表示。
该法案出台之际，性能愈发强大的人工智能模型正对国家安全和公共安全构成风险。6月12日，美国商务部以国家安全为由对Anthropic公司的最新模型采取行动，导致Anthropic在全球范围内暂停了这些模型的访问权限。政府的这一指令暴露了缺乏透明的前沿人工智能监管框架的问题。
该草案法案规定需上报的活动包括：人工智能模型试图规避人类监督、绕过安全防护措施，以及以其他方式破坏人类操作员对模型的控制能力。上报范围还包括未经授权访问模型权重（该参数决定人工智能系统的决策逻辑），以及对公共安全构成威胁的化学、生物、核和其他风险。
该法案是国会提出的最新一项人工智能监管法案。此前，国会就联邦法律是否应优先于州级法律、监管措施是否会减缓创新以及美国与中国的竞争等问题展开辩论，导致立法进程陷入停滞。
本月早些时候，众议院的两名议员公布了一项广泛的人工智能立法讨论草案，即《伟大美国人工智能法案》，其中也包含了向商务部上报重大安全事件的条款。
莫兰表示，他提出的更具针对性的法案有望更快获得通过，并认为该法案将迅速获得两党支持。
“此前几乎没有任何人工智能立法有机会通过，但我认为公众要求采取行动的呼声越来越高，”支持莫兰法案的人工智能政策网络主席马克·贝尔说道。
记者：卡伦·弗赖菲尔德
编辑：千住典子和大卫·加芬
本报守则：路透社信托原则。
US lawmaker introduces bill to require AI companies to report critical incidents
2026-06-25 2:03 PM UTC / Reuters
By Karen Freifeld
June 25, 2026 2:03 PM UTC Updated 19 mins ago
Congressman-elect Nathaniel Moran (R-TX) attends orientation for newly elected members of the 118th U.S. Congress, following the 2022 midterm elections, outside the Hyatt Hotel in Washington, U.S., November 13, 2022. REUTERS/Michael A. McCoy
Summary
- Bill would require Commerce Department notification
- Comes amid growing calls for AI regulation as models become more powerful
- Targeted bill hopes to break through Congress’s struggle to pass AI legislation
June 25 (Reuters) – A Republican lawmaker on Thursday proposed legislation that would require AI model developers to report dangerous capabilities, security breaches and safety incidents.
The draft legislation, introduced by U.S. Representative Nathaniel Moran of Texas, would mandate AI companies to report to the U.S. Commerce Department within seven days of discovering dangerous activity, with Commerce required to notify Congress within 48 hours of the most serious incidents.
Jumpstart your morning with the latest legal news delivered straight to your inbox from The Daily Docket newsletter. Sign up here.
“It’s a catch-it-early and sound-the-alarm bill,” Moran said in an interview about the AI Incident Reporting Act.
The bill comes as ever-more powerful AI models raise the risks to national security and public safety. On June 12, the Commerce Department took action against Anthropic’s latest models in the name of national security that resulted in Anthropic disabling access to them globally. The government directive exposed the absence of a transparent framework to govern frontier AI.
Reportable activity under the draft legislation includes a model attempting to evade human oversight, circumvent safeguards, and otherwise undermine the ability of human operators to control the model. It also includes unauthorized access to model weights, which help determine a machine’s decision-making, and chemical, biological, nuclear, and other threats to public safety.
The bill is the latest AI regulation to be proposed in Congress, which has struggled to pass legislation amid debate over whether federal law should preempt state laws and whether innovation and the U.S. competition with China would be slowed by guardrails.
Earlier this month, two lawmakers in the House of Representatives released a discussion draft of broad AI legislation known as the Great American Artificial Intelligence Act, which included reporting critical safety incidents to Commerce.
Moran said his more targeted approach could find a quicker path to law, and thought it would bring in bipartisan support quickly.
“No legislation on AI has had much of a chance, but I think there’s a growing demand from the public to see some action,” said Mark Beall, president of the AI Policy Network, who supports Moran’s proposed legislation.
Reporting by Karen Freifeld; Editing by Chizu Nomiyama and David Gaffen
Our Standards: The Thomson Reuters Trust Principles.
发表回复