Technology
Anthropic pitches AI safety bill with $1 billion trigger
The Hindu BusinessLine··7h ago
Anthropic representatives pitched US lawmakers a draft frontier AI safety bill, Politico reported on Friday. Officials could block a model only for imminent risk above $1 billion in damages, or serious harm to more than 50 people. Independent evaluators would register with the Commerce Department and report any severe risks to it within 72 hours.
Prism
What It Means For You
- The draft sets a high damage and casualty threshold before federal officials could halt a frontier model release.
- Labs could pay independent evaluators if payment is not tied to the evaluation outcome, per the report.
- A federal judge in the District of Columbia could order a model suspension after a Commerce risk finding.
What's Happening
- Covered risks include biological weapons, a model evading developer controls, or a cyberattack, Politico said.
- An Anthropic spokesperson told Reuters the text tracks the firm's Advanced AI Framework.
- The Department of Commerce did not immediately comment to Reuters.
How Enforcement Would Work
- Evaluators finding imminent enumerated risk must notify Commerce within 72 hours of finishing an assessment.
- The language bars other US agencies from suspending development or internal use outside that path.
- Politico cited three people familiar with the Capitol Hill pitching effort.
all-news
