Anthropic Draft Bill Sets High Bar for Blocking AI Models
The proposal would allow intervention over specified risks and require independent evaluations of frontier AI models.
Key takeaways
- The draft sets thresholds of more than $1 billion in damages or harm to over 50 people.
- Independent evaluators would review frontier models and report certain imminent risks to Commerce.
- A federal judge could order a suspension after the Commerce Department seeks one.
Anthropic has pitched US lawmakers a draft bill on AI safety and security that would let the government seek to block a model’s release or use in narrowly defined cases, according to Politico.
The proposal sets a threshold of a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. The risk would have to stem from a biological weapon, a model evading its developer’s controls or a cyberattack.
Anthropic’s spokesperson said the bill reflects proposals in the company’s Advanced AI Framework, which it has advocated for with federal and state lawmakers since June. The draft would require developers to test frontier models, share findings and allow independent evaluators to review the models.
According to Politico, evaluators would register with the Commerce Department and report an “imminent enumerated risk” within 72 hours of completing an assessment. The department could ask a federal judge in Washington to suspend a model; other government departments and agencies would be barred from doing so.
The proposal would also override some state laws, according to Politico. The company declined to comment on how widely it had presented the draft or lawmakers’ response, The Business Times reported.
This article was produced with the help of AI technology. Based on reporting by the outlets listed above.
