OpenAI backs watered-down Massachusetts AI safety bill - Politico

OpenAI on Tuesday endorsed a Massachusetts state bill that would establish new safety rules for AI development, but which falls short of a stricter House proposal backed by Anthropic, according to Politico. OpenAI's endorsement signals its preference for a lighter regulatory touch, contrasting with Anthropic's support for a more aggressive House version that would require mandatory testing and licensing for advanced AI models.

However, it does not mandate pre-deployment testing or third-party audits, which are key features of the House version. Both bills remain in committee, and the legislative session is scheduled to end in July 2028.

The endorsements come amid a broader push by states to regulate AI, with at least 15 states considering similar legislation this year. Lawmakers in Massachusetts have been debating AI oversight since 2024, following a series of well-publicized incidents involving model misbehavior.