What Happened
The AI liability insurance market just split in two directions at once.
On one side, Armilla AI - now the first Lloyd's of London coverholder dedicated exclusively to AI liability - expanded its standalone AI Liability Policy to cover up to $25 million per organization. The policy, underwritten by Chaucer and other Lloyd's syndicates, explicitly covers what traditional insurers won't touch: hallucinations, model drift, inaccurate outputs, data leakage, and AI regulatory violations. It also covers claims tied to harmful outputs like defamation, trade secret exposure, and confidentiality breaches. This is purpose-built insurance for companies deploying AI in production.
On the other side, Verisk rolled out new exclusion endorsements effective January 1, 2026, giving traditional carriers the option to exclude generative AI from general liability policies entirely. Three specific ISO forms (CG 40 47, CG 40 48, and CG 35 08) let insurers drop coverage for bodily injury, property damage, or personal injury "arising out of generative artificial intelligence." Verisk defines generative AI broadly as any "machine-based learning system or model trained on data with the ability to create content." Carrier interest in adopting these exclusions has been strong, and because Verisk's templates get widely adopted across the US insurance industry, the effect could ripple through quickly.
Meanwhile, US states are adding regulatory teeth. California's SB 243 now requires AI disclosure and safety protocols for companion chatbots. The Texas Responsible AI Governance Act (TRAIGA) took effect. The EU AI Act enforcement continues ramping up with heavy fines for high-risk AI systems.
Why It Matters
If your company uses ChatGPT, Claude, or any AI tool in customer-facing workflows, your existing business insurance may soon explicitly exclude AI-related claims. That's not a hypothetical - it's happening right now with the Verisk endorsements.
This creates a real coverage gap. A chatbot hallucinating medical advice, a coding assistant introducing a security vulnerability, an AI generating defamatory content about a competitor - these scenarios were previously in a gray area under general liability. Now carriers can cleanly exclude them.
For smaller businesses and startups, this is particularly sharp. Specialized AI liability coverage from players like Armilla exists, but it's an additional cost that most haven't budgeted for. And the $25 million cap, while substantial, may not cover catastrophic failures at scale.
For AI tool vendors themselves, this changes the competitive landscape. Tools that can demonstrate lower hallucination rates, better guardrails, and auditable outputs become more insurable - and their customers become more insurable too.
Our Take
This is the clearest sign yet that AI deployment risk is being priced by the market, not just debated in policy papers. Insurance actuaries don't care about hype cycles. When they start writing exclusions and specialized policies, it means the risk data is real enough to model.
The two-track approach makes sense but creates winners and losers. Companies that proactively get AI liability coverage gain a competitive advantage - they can tell enterprise clients "yes, we're insured for AI failures." Companies that ignore this will eventually face a rude surprise when a claim gets denied under a newly-adopted Verisk exclusion.
For AI tool users, the practical takeaway: check your existing policies for AI exclusions, especially after renewal. Document your AI governance practices. And when evaluating tools, start asking vendors about their own liability coverage. The era of "move fast and deploy AI" without thinking about liability is closing.
The regulatory patchwork across US states makes this messier than it needs to be. But the insurance market is doing what regulators have been slow to do - putting a real price tag on AI risk.