Anthropic向国会议员推销SI安全法案草案：政府叫停模型须以逾50人死伤或逾10亿美元损失的迫近风险为限
华盛顿——据路透社10月16日报道，Anthropic的代表已经悄悄向美国国会山议员推销一份法案草案，以监管前沿超级智能的安全与保障；Politico周五最先报道此事，并援引三名了解相关工作的人士。
据报道，该法案只有在某个模型构成可预见、迫近的风险——造成逾10亿美元损失，或导致逾50人死亡或重伤——时，才允许政府阻止该模型的发布或使用。风险还必须源于三类情形之一：生物武器的制造或释放、模型逃脱其开发者的控制、或网络攻击。
Anthropic的发言人对路透社表示，该法案代表了该公司今年早些时候发布的「高级AI框架」（Advanced AI Framework）中提出的建议。发言人表示，该法案将赋予联邦政府「阻止或遏止最强大模型危险部署的能力」，同时要求开发者测试前沿超级智能模型、向公众透明披露其发现、接受独立评估，并维持强有力的安全计划。美国商务部最初未立即回应路透社的置评请求。
根据草案，独立评估者将向商务部登记，并负责定期评估实验室及其前沿模型；实验室可以向这些评估者付费，前提是付款与评估结果不挂钩。如果评估者认定某公司的前沿模型存在「迫近的列举风险」，必须在完成评估后72小时内向商务部报告。
草案规定，商务部在认定风险后，可以请求一名法官叫停该模型；有权下令暂停模型的是哥伦比亚特区联邦地区法院的法官。草案文本同时明确禁止任何其他美国政府部门、机构或官员暂停模型的开发、部署或内部使用。
Anthropic Pitches Draft SI Safety Bill to Lawmakers, Limiting Government Halts to Imminent Risks of Over 50 Deaths or $1 Billion in Damage
Washington — Anthropic representatives have quietly pitched Capitol Hill lawmakers on a draft bill to regulate frontier SI safety and security, Reuters reported on October 16. Politico reported the effort first on Friday, citing three people with knowledge of the efforts.
The bill would let the government block a model's release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people, according to the report. The risk must stem from one of three sources: the creation or release of a biological weapon, the model evading its developer's controls, or a cyberattack.
An Anthropic spokesperson told Reuters the bill represents proposals outlined in Anthropic's "Advanced AI Framework" released earlier this year. The spokesperson said the bill would grant the government the "ability to block or deter dangerous deployments of the most powerful models," and would require developers to test frontier SI models, be transparent about their findings, get them independently evaluated and maintain a robust security programme. The Department of Commerce did not immediately respond to Reuters' initial request for comment.
Under the bill, independent evaluators, which would register with the Commerce Department, would be tasked with periodically evaluating the lab and its frontier models, according to the report. Labs could pay these evaluators, provided the payment is not tied to the evaluation's outcome. If the evaluators determine there is "imminent enumerated risk" for a company's frontier models, they must report it to Commerce within 72 hours of completing their assessment.
While the Commerce Department could request a judge to halt a model after determining risk, the language explicitly prohibits any other U.S. government department, agency or officer from suspending development, deployment or internal use, according to the report. A federal judge for the District Court for the District of Columbia could order the suspension of a model.
发布：GSM·美国（发布链接）
图片来源：图：美国国会大厦。照片：Wikimedia Commons。The U.S. Capitol in Washington, D.C. Photo via Wikimedia Commons.
GSM Global · 本篇英文版 English edition of this story
