Anthropic Warns AI Risks May Be Uninsurable in IPO Filing
Anthropic disclosed in its IPO filing that existing insurance may not cover catastrophic AI incidents, highlighting a systemic gap in coverage for physical AI-related harm.
AI developer Anthropic disclosed in its S-1 IPO filing that its current insurance policies may be insufficient to cover liabilities resulting from catastrophic AI-related incidents. The company warned that the scale of potential harm could exceed the capacity and expertise of any single insurer or the insurance industry in aggregate, noting that models may behave in ways that no user or developer intended.
This disclosure underscores a broader industry struggle to define AI coverage. While current claims focus on data privacy and intellectual property theft, standard cyber insurance policies typically exclude bodily injury and physical harm. Research from Gartner suggests this creates a gaping hole in coverage, predicting that claims involving death caused by AI could exceed 2,000 worldwide by the end of 2026.
Some firms are attempting to address these gaps. Armilla Insurance Services launched a liability policy protecting against hallucinations and unexpected AI behavior, while Universal Underwriters Insurance Company implemented a suite of approved AI tools to manage risk. Additionally, Google partnered with insurers to shield Cloud customers from trade secret losses caused by rogue AI. Despite these efforts, Zurich North America noted that different lines of business are handling the transition differently.
