Generative AI is rapidly embedding itself into enterprise operations, from customer service bots to automated legal drafting tools. Large language models (LLMs) are now integral to core business functions. But with widespread adoption comes unprecedented legal and financial exposure: hallucinated facts, misleading advice, and inappropriate outputs have already caused costly incidents, and traditional insurance frameworks often fall short.

In response to potential multi-billion-dollar claims, major U.S. insurers including AIG, Great American, and WR Berkley have sought regulatory approval to exclude AI-related liabilities from corporate policies. These exclusions cover not only in-house AI deployments but also products or services marketed as AI-enabled. Rather than retreating from the market, insurers are aiming to manage risk exposure while keeping options open for future coverage.

At the same time, some insurers are developing highly targeted AI policies. Hiscox, Chubb, Beazley, and Munich Re are among the first to offer coverage for losses caused by AI errors, including legal defense costs, contractual failures, and reputational harm. Compared with traditional E&O or cyber liability policies, these products incorporate more nuanced criteria: distinguishing between model-level errors, prompt-engineering mistakes, and human oversight lapses, while excluding malicious or unauthorized use. Policies often require enterprises to maintain complete audit trails, prompt logs, and human-in-the-loop review processes.

Recent incidents highlight the stakes: a Google AI snippet falsely accused Wolf River Electric of being investigated by state authorities, triggering a defamation lawsuit exceeding $110 million; a Canadian airline had to honor discounts generated by its chatbot; and UK engineering firm Arup lost HKD 200 million after a fraudster exploited a digital executive clone. As Kevin Kalinich of Aon notes, while single-company losses can be absorbed, model-level failures that trigger thousands of simultaneous claims pose systemic risks insurers cannot underwrite alone.

For AI firms, traditional insurance may not cover all potential liabilities. OpenAI and Anthropic, facing high-profile copyright and safety lawsuits, are considering using investor funds to supplement insurance coverage. OpenAI reportedly holds approximately $300 million in insurance via Aon—far below potential multi-billion-dollar payouts—while Anthropic will use internal funds for part of its $1.5 billion settlement obligations.

Despite these challenges, insurers' growing engagement with AI is fostering more robust governance. AI liability coverage is not just a risk-transfer mechanism; it is shaping enterprise AI oversight. In the future, insurance may be bundled with AI platforms, incorporating risk-scoring models, prompt and output monitoring, and human oversight to provide end-to-end liability management.

The emergence of AI-specific insurance marks a turning point: generative AI has transitioned from an experimental tool to a core enterprise function. The insurance industry is not merely adapting to new risks—it is helping define the compliance and safety boundaries of AI deployment, establishing the frameworks that enterprises will need to manage technology responsibly.