Anthropic 向美国议员推销自拟 AI 安全法案
Anthropic quietly pitches its own AI safety Bill to US lawmakers
Anthropic 自己写了个 AI 安全法案拿给美国议员,门槛定得特别高:要损失超 10 亿美元或伤 50 人才拦,很多公司都达不到这标准。
Anthropic 向美国议员提交了一份自行起草的 AI 安全法案。法案规定,政府只有在模型造成可预见、迫近的风险时才能阻止其发布或使用。风险门槛设定为超过 10 亿美元损失,或造成 50 人以上死亡或重伤。该门槛明显偏高,意味着绝大多数模型不会受到发布限制。
Anthropic quietly pitches its own AI safety Bill to US lawmakers
The Bill would reportedly let the government block a model's release or use only if it poses a foreseeable, imminent risk of more than US$1 billion in damages, or of killing or seriously injuring more than 50 people.