Insurers balk at multibillion-dollar claims faced by OpenAI and Anthropic (www.ft.com)

Reports say major insurers are refusing or limiting coverage for the multibillion-dollar liability claims reportedly targeting OpenAI and Anthropic, leaving those companies exposed to large legal and regulatory payouts. Insurers and reinsurers are balking because AI liability is novel, potentially systemic, and hard to price: policies weren’t written for algorithmic harms that can scale rapidly, causation is unclear, and historical loss data are sparse. As a result, carriers are carving out AI-related exposures, pushing higher premiums, or demanding narrower terms — moves that could force firms to self-insure, buy bespoke programs, or slow product rollouts. This standoff matters for the AI community because insurance is a critical risk-management lever for startups and established labs alike. Tight or costly coverage raises the cost of deploying models, could chill innovation, and shifts more regulatory and financial risk back onto developers. Technically, it underscores the need for better empirical risk measurement (model auditing, provenance logs, incident forensics), stronger safety engineering (robustness, interpretability, access controls), and clearer legal definitions of harm and liability. Expect growth in tailored “AI risk” insurance products, closer collaboration between insurers and ML teams on loss-prevention standards, and potential policy intervention to clarify coverage for emergent AI risks.