Anthropic Quietly Pitches Its Own AI Safety Bill to Lawmakers, Politico Reports

3-line summary

• Anthropic representatives have quietly proposed draft legislation to Capitol Hill lawmakers to regulate frontier-AI safety and security, Reuters reported citing Politico, which cited three sources familiar with the matter.
• The draft would let the government block a model's release or use only for foreseeable and imminent risks meeting a threshold of over $1 billion in damages or 50-plus deaths/serious injuries, limited to three risk types: bioweapon creation or release, a model escaping developer control, or cyberattacks; independent evaluators registered with the Commerce Department would periodically assess labs and frontier models, with lab-paid fees barred from being tied to evaluation outcomes.
• Evaluators must report an imminent enumerated risk to Commerce within 72 hours of completing an assessment; Commerce could then seek a court order in Washington D.C. federal district court to halt a model, while other federal agencies would be barred from independently halting development, deployment, or internal use. Anthropic and Commerce did not immediately respond to Reuters' request for comment.

This summary is our own processing of the source article; see the source link below for full context.