Anthropic Submits AI Safety Bill Draft to US Congress, Proposing Government Power to Block High-Risk Models
SingTao · 2 SOURCESabout 2 hours ago2 MIN
Summary
Anthropic, an AI startup, has privately presented a draft bill to US lawmakers that would establish a new framework for regulating artificial intelligence safety. The proposed legislation would authorize the government to block or halt AI model deployments under specific circumstances, including models that pose imminent threats capable of causing more than 50 deaths or injuries, or economic losses exceeding $1 billion. The company states the bill reflects recommendations from its "Advanced AI Framework" and has been advocated to federal and state legislators since June.
Key Points
- The bill authorizes government intervention only when AI models create foreseeable urgent risks resulting in over $1 billion in losses or 50-plus casualties
- Intervention triggers are limited to three scenarios: development of biological weapons, models escaping developer control, or large-scale cyberattacks
- AI developers would be required to conduct mandatory security testing and publish evaluation results for public transparency
- Independent evaluation agencies must conduct regular reviews of AI laboratories, with urgent risks reported to the Commerce Department within 72 hours
- The Commerce Department can petition judges to order model shutdowns if risks are confirmed after evaluation
Why It Matters
The proposed legislation represents one of the most concrete regulatory frameworks advanced by a major AI company, potentially setting a precedent for how governments worldwide address catastrophic AI risks . Recent AI-powered security breaches and cyberattacks have heightened concerns about potentially destructive applications, making such regulatory frameworks increasingly urgent despite the Trump administration's apparent reluctance to strengthen oversight .
The proposed legislation represents one of the most concrete regulatory frameworks advanced by a major AI company, potentially setting a precedent for how governments worldwide address catastrophic AI risks . Recent AI-powered security breaches and cyberattacks have heightened concerns about potentially destructive applications, making such regulatory frameworks increasingly urgent despite the Trump administration's apparent reluctance to strengthen oversight .