Science & Tech · AI · 17 mins ago
Anthropic proposes bill to regulate risks from frontier AI
Anthropic has proposed a bill to regulate risks from frontier AI.
Frontier AI refers to advanced AI systems, the type of technology the proposal aims to address.
The bill would require independent evaluators to report risks they consider imminent.
They would have 72 hours after an assessment to make that report.
This would set a time limit for alerting others to urgent risks identified in an evaluation.
The proposal does not say what would happen after a risk is reported.
Anthropic has proposed legislation setting limits on when federal authorities could halt advanced AI models.
The draft would allow intervention only for a foreseeable, imminent threat involving more than $1 billion in damages or the deaths or serious injuries of over 50 people.
It would require the Commerce Department to seek a federal judge’s order before suspending a model.
Independent evaluators would assess developers and notify Commerce of covered imminent threats within 72 hours of completing an assessment.
The proposal would also override some state requirements and enters a broader debate over federal oversight of frontier AI.
- Who
- Anthropic proposed the draft bill. It would apply to frontier AI developers and federal authorities.
- What
- The bill would set rules for oversight and government intervention involving frontier AI models.
- When
- The articles were published on October 17, 2026. Anthropic says it has promoted its Advanced AI Framework since June.
- Where
- The proposal concerns US federal oversight and names the federal District Court for the District of Columbia.
- Why
- Anthropic says the legislation reflects its Advanced AI Framework. It would give developers a clearer legal framework and limit when the government can intervene.
Anthropic
Safety advocates and lawmakers seeking stronger federal review or broader restrictions
Government intervention
Anthropic
Anthropic’s draft would limit intervention to specified imminent catastrophic threats and require a federal judge’s order to suspend a model.
Safety advocates and lawmakers seeking stronger federal review or broader restrictions
Some safety advocates and Senator Maria Cantwell want a stronger federal review process.
Restrictions on advanced AI
Anthropic
Anthropic’s proposal takes a narrower approach, setting safety and reporting requirements for developers.
Safety advocates and lawmakers seeking stronger federal review or broader restrictions
Representative Greg Casar and Senator Bernie Sanders have called for pausing advanced AI development and banning superintelligence.
Anthropic has promoted its Advanced AI Framework among federal and state lawmakers.
US export controls reportedly sidelined two advanced Anthropic models.
Articles reported on Anthropic’s proposed bill to regulate frontier AI risks.
- Damage threshold
- More than $1 billion
- Casualty threshold
- Deaths or serious injuries of over 50 people
- Evaluator notification deadline
- Within 72 hours of completing an assessment
- Court named in draft
- Federal District Court for the District of Columbia
- Permitted threat scenarios
- Biological weapons, models escaping developers’ controls, or cyberattacks
3 sources
What people are saying
These are sensible policies we believe Congress should act on