Anthropic 的人工智能安全法案草案仅允许 Commerce 寻求暂停发布

该草案规定,只有在造成超过50人死亡或重伤,或造成超过10亿美元损失的情况下,商务部才能请求法院停止发布。

By · Published · Updated

Primary source: Tom's Hardware

为何重要

Anthropic 正在尝试将外部模型评估纳入正式的政府程序。该草案设定的较高危害门槛以及赋予 Commerce 的独占角色将决定哪些风险可以触发干预,以及哪个机构可以寻求阻止发布。

Anthropic's AI safety bill gives Commerce the only release-stop button — The draft would require an evaluator to flag imminent risks before Commerce could ask a federal judge to block a model release.

Anthropic CEO and co-founder Dario Amodei 正在呼吁立法者为最危险的 AI 发布设立狭窄的法律门槛:根据一份在 10 月 16 日报道的草案,该政府只有在模型构成对超过 50 人被杀或重伤的迫在眉睫风险,或造成超过 $1 billion 的损失时,才可寻求阻止该模型的发布。据 Tom's Hardware 对 Politico 报道的转述,该提案将使 Commerce…

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

Reader comments

Conversation for this story loads after sign-in.