Anthropic briefs lawmakers on draft frontier model bill with release-block powers

Anthropic representatives have pitched a draft bill on frontier AI safety and security to Capitol Hill lawmakers, according to Reuters, which cited a Politico report and three people familiar with the effort. As described, authorities could block the release or use of a frontier model only where risks are imminent and significant. The named risk areas include biological weapons, cyberattacks, and cases where a model evades its developer's control mechanisms. The draft also proposes independent evaluators registered with the Department of Commerce, who would test models periodically and report urgent risks. The approach gives the developer's own safety work a formal outside check, which differs from voluntary commitments. For founders building on frontier models, the proposal matters because it would define when a provider can withdraw a model and who verifies its safety. Teams that depend on third-party evaluations should track whether Commerce-registered evaluators become a compliance requirement.