Anthropic IPO filing exposes a gap in AI insurance coverage

Anthropic’s recent IPO filing reveals a growing tension between the rapid expansion of artificial intelligence and the lagging development of insurance products designed to cover it. The company, which holds more cash than almost any other firm in the United States, admitted in its S-1 document that it cannot confidently answer whether existing policies will adequately cover potential liabilities.

The filing states that models may behave in unintended ways, creating substantial liability. Anthropic noted that its insurance policies might not provide sufficient coverage, especially when a single incident causes disproportionate harm. The company warned that no single insurer, or potentially the entire industry, may have the capacity to underwrite the exposure from a catastrophic AI-related event.

Alissa Lugo, a senior director analyst at Gartner, estimates that claims involving AI in fatalities will exceed 2,000 worldwide by the end of 2026. She defines these as cases where AI played some role in a death. Lugo observes that more companies are adding AI exclusions to their policies, leaving courts to define the parameters of coverage in the future.

Dennis Hendrickson of Hendrickson Insurance in Sarasota, Florida, points out that cyber insurance currently covers most AI-related financial losses. However, he argues that cyber policies were built for data breaches and system outages, not for physical harm. He describes the lack of bodily injury coverage as a gaping hole that will need to be addressed as AI becomes more integrated into business operations.

Kara Higginbotham of Zurich North America explained that insurers are wary of writing AI coverage because the risks are new and poorly understood. She stated that carriers generally do not cover bodily harm from AI incidents by default. Underwriters often ask companies to detail where they use AI and who oversees it, but many businesses cannot yet answer that question clearly.

Vikram Desai of Accenture compared the current situation to the early days of cybersecurity. He noted that cyber risk initially sat inside general liability policies before insurers named it and excluded it. Desai warned that AI is following the same curve but at a compressed pace, creating exposure that companies may not realize they have.

Jaymin Kim of Marsh’s cyber practice argued that AI is not a new peril but a technology through which existing harms manifest. She suggested that separating AI into its own insurance category could be tedious and impractical. Kim maintained that most current cyber policies are sufficient to handle AI issues, though she conceded that the industry has not yet seen a wholly new category of risk created by the technology.

Hendrickson predicts that insurers will eventually set the standards for AI use before the government does. He expects a wave of claims that will force carriers to tighten underwriting and lower sub-limits. He warned that independent agents will bear the brunt of dealing with uncovered customers when policies are not renewed after a claim is filed.