Anthropic Warns Catastrophic AI Harm Could Exceed Insurance Capacity

Anthropic's Oct. 26 IPO filing warns a catastrophic artificial intelligence incident could exceed the capacity or expertise of one insurer and potentially the industry as a whole, raising questions about coverage for AI-related losses.

Anthropic filed its Form S-1 Monday, Oct. 26, for a proposed initial public offering. The filing says its models may behave in ways users or developers do not intend, potentially exposing the company to substantial liability. Existing policies may not adequately reduce that risk, particularly if one event or a related group of events causes disproportionate harm.

The company describes AI-specific insurance as a market still taking shape and warns it may be unable to obtain coverage for some AI liabilities on commercially reasonable terms or at all. The S-1 gives no dollar estimate for the potential exposure and does not report an insurance denial; it outlines possible future risks.

The S-1 also details contractual risks tied to testing. Some agreements with external evaluators require Anthropic to defend and indemnify them without a liability cap for specified third-party claims alleging infringement involving its models or outputs during authorized testing.

For other businesses, an AI-related claim could involve a data breach, faulty output, intellectual-property dispute or physical injury. These harms may fall under different insurance lines, so whether a policy responds depends on its wording and the specific loss.

Insurers have introduced endorsements to clarify or expand AI coverage in cyber and technology errors-and-omissions policies. Specialized liability products, including difference-in-conditions contracts, are also available. Meanwhile, general-liability exclusions available beginning Jan. 1, 2026, may remove coverage for some generative AI-related bodily injury or property damage claims when a policy adopts them; their availability does not mean they apply to every policy. Marsh opposes broad or absolute AI exclusions that it says could create gaps in client protection.

For companies using AI, the filing leaves a policy-specific question: whether their coverage applies to the particular harm an AI system may cause.