Anthropic, OpenAI Split Over Massachusetts AI Safety Bill

Massachusetts is moving to fill gaps in US federal oversight with what could become the country’s toughest state-level artificial intelligence safety regime. The proposal targets potential risks from advanced AI models and seeks greater transparency from developers. If enacted, it could influence legislation elsewhere and establish a consequential compliance benchmark for leading AI companies operating across state lines.

The bill would require AI models to undergo an assessment by an outside organization every 120 days, with the findings made public. Anthropic and OpenAI have split over that evaluation mechanism, highlighting disagreement among the industry’s leading developers over the frequency of reviews, the role of external assessors and the scope of public disclosure, even as both companies engage with the broader push for AI safety rules.