OpenAI and Google resist Massachusetts frontier AI safety rules that Anthropic backs

OpenAI and Google are opposing Massachusetts language that would require leading AI developers to undergo independent catastrophic-risk evaluations of frontier models on a four-month cycle, a provision Anthropic supports as the nation’s strictest state-level AI safety standard. The split, unfolding as lawmakers reconcile House and Senate versions of a broader economic development bill, is being watched as a template for how other states might police the most capable models.