Anthropic IPO filing reveals insurers may lack capacity for catastrophic AI failures

Anthropic’s S-1 filing identifies a potential gap in the insurance market that could leave the company exposed to substantial liability. The document states that no single insurer, and possibly not the insurance industry as a whole, may have the capacity or expertise to underwrite coverage matching a catastrophic AI-related incident.

The filing details specific risks, including model behavior that differs from user or developer intent, data loss, and security incidents. It warns that a single incident or a cluster of incidents could cause disproportionate harm that existing policies do not fully address. Anthropic notes the market for AI-specific risks is nascent, meaning it may be unable to obtain some coverage on commercially reasonable terms, or at all.

These concerns extend beyond large technology firms to any organization using AI tools. Alissa Lugo, a senior director analyst at Gartner, says her research indicates that “death by AI” claims, defined as cases where AI played some role in a fatality, could exceed 2,000 worldwide by the end of 2026. She observes that more companies are adding AI exclusions to their policies and that insurers, with help from courts, will set the parameters of coverage.

Insurance brokers describe the current landscape as fragmented. Dennis Hendrickson of Hendrickson Insurance in Sarasota, Florida, says cyber policies currently provide the most coverage for AI, but only for specific losses such as data breaches, privacy violations, and deepfake wire fraud. He calls the exclusion of bodily injury a “gaping hole.” Kara Higginbotham of Zurich North America agrees that a gap exists for physical harm, noting carriers are wary of writing coverage for scenarios that remain difficult to underwrite.

Henrietta Akpata of Marist University points out that insurers are telling clients, “We can’t cover what you can’t show us.” She notes that most companies cannot accurately state where employees use AI tools. This lack of visibility creates exposure for local businesses, such as those using website chatbots or AI for hiring, which may face problems that are no longer covered by standard policies.

Experts disagree on the future trajectory of AI coverage. Hendrickson expects financial-loss coverage to become routine and priced similarly to cyber insurance. He predicts bodily injury will face tight underwriting, low sub-limits, and requirements for human review before AI output reaches a patient. Jaymin Kim, who leads emerging technologies at Marsh, argues that most current cyber policies are sufficient. “AI is not a peril,” she said, describing it as a technology through which familiar harms appear. She conceded that “maybe the operative word is ‘yet.’”

Vikram Desai, a global cyber strategy lead at Accenture, compares the current situation to the early days of cyber risk. He notes that cyber risk sat silently inside general liability for years until insurers named it, excluded it, and then sold it back as a separate product. He says AI is following the same curve in a compressed form. Desai adds that underwriters are already asking where AI is used and who oversees it, with exclusions applying when AI sits anywhere in the causal chain.

Higginbotham states that most AI-related claims today are routine matters, primarily intellectual property disputes against developers accused of scraping copyrighted content, plus data misuse and privacy claims. She says pricing could rise if claim patterns shift in a sustained way, but “we are not there yet.” Hendrickson, meanwhile, expects a wave of claims that no one is prepared for and predicts insurers, not government, will set AI standards for many businesses.