2026-08-20 · Source: Insurance Journal
Summary in 3 Points • Massachusetts lawmakers propose stringent AI safeguards, requiring independent reviews of AI models every 120 days • OpenAI warns these reviews could slow cybersecurity model releases, while Anthropic supports the proposal • The legislation is part of a broader economic bill likely to pass before the November election --- Massachusetts is poised to implement the strictest state-level AI regulations in the US, requiring independent evaluations of AI models every 120 days to assess catastrophic risks. This proposal, included in an economic development bill passed by the state Senate, aims to ensure AI safety while fostering the state's AI sector. OpenAI has expressed concerns that the frequent reviews could delay the release of cybersecurity models, advocating for a uniform standard similar to Illinois. In contrast, Anthropic supports the Massachusetts proposal, arguing for more stringent oversight. The bill, which also includes funding for local projects, is expected to pass before the November elections, with final details being negotiated by a group of House and Senate members.
The Massachusetts AI legislation could influence global regulatory trends, potentially affecting underwriting and risk assessment in the London Market. Insurers may need to consider the implications of more frequent AI model evaluations on product development timelines and risk profiles. The proposed safeguards could lead to increased demand for AI-related liability coverage as companies navigate compliance. Additionally, the focus on catastrophic risk assessment may prompt insurers to reevaluate exposure and policy wording related to AI technologies.