The conversation happening in broker offices across the US right now goes something like this: a client asks whether their business is covered if their AI system causes harm. The broker checks with the carrier. The carrier declines to confirm. Everyone moves on.
That gap - silent, unpriced, and unaddressed - is now a live coverage problem. As the litigation picture around generative AI sharpens, the standard market is moving to exclude the exposure from commercial general liability (CGL), errors and omissions (E&O), and professional liability policies.
According to Gallagher Re, generative AI-related lawsuits in the US grew 978% between 2021 and 2025, with year-over-year filing increases accelerating to 137% in 2024-2025. As litigation climbs, carriers are rewriting the terms at renewal. Major carriers including Chubb, Travelers, and Berkshire Hathaway filed for, and in most cases received, state regulatory approval to add AI exclusions to commercial general liability policies. W.R. Berkley went further, filing an absolute AI exclusion - form PC 51380 - designed to strip coverage across D&O, E&O, and fiduciary lines for any claim connected to AI use. ISO introduced multiple generative AI exclusion endorsements for its CGL coverage forms effective January 2026.
The result, says George Lewin-Smith (pictured), CEO and co-founder of Testudo - a Lloyd's coverholder writing standalone generative AI liability insurance for US businesses - is that most commercial clients are sitting on an uninsured exposure they don't yet know exists.
"The brokers are not particularly willing to say, 'Oh yeah, you're absolutely covered for this generative or agentic risk in your business,'" Lewin-Smith told Insurance Business. "They're not willing to get that clarification or clarity, black and white, by the insurers, because the insurers are saying, 'I had no idea I have all this silent exposure in my book. I have no intention of covering agents or AI in my GL. I'm not pricing for it, I'm not accounting for it.'"
Read more: AI exclusions are spreading through commercial policies, brokers warn clients to check cover
The source of confusion for most commercial buyers is that AI liability doesn't map cleanly onto any existing policy type.
Cyber insurance is designed for data breach scenarios - when a company is attacked and information is extracted. When an AI system produces harmful or incorrect output without any external breach, cyber doesn't respond. "No-one has really breached you," Lewin-Smith said. "No-one's tried to extract data from you."
E&O and professional liability create a different problem. Law firms, construction companies, and financial services firms using AI tools are not technology companies. They generally don't buy tech E&O, which means they face a gap: their professional liability wasn't designed for AI-generated output, their GL is now being excluded, and their cyber policy doesn't apply.
If a general liability policy excludes generative AI, brokers need to confirm whether the client's E&O or cyber policy picks up the exposure, or whether it falls between policies.
Lewin-Smith is direct about where responsibility sits in these scenarios. The vendor contracts and base model provider terms of service, he says, are unambiguous: the deployer of the AI technology owns the liability for its outputs. "You pick up a sword and you swing the sword around, you're responsible for the outcome, despite not selling the sword,” he said.
Banks, law firms, airports, and healthcare companies using AI in their operations are the businesses most exposed - and least likely to have considered whether their current program covers the risk.
The practical challenge for brokers is that the exclusion landscape is uneven and moving fast. Under CG 40 47, an insurer may deny coverage if the loss arose from generative AI - the trigger is the connection to AI, rather than the proportion of work it touched.
That matters for clients where AI is one tool among many: a law firm using an AI drafting platform, a lender running automated credit assessments, or a hospital system relying on AI-assisted diagnostic tools. If a claim has any AI connection, the exclusion may void coverage entirely.
There is another market dynamic worth watching. Lewin-Smith points out the tension between brokers publicly resisting AI exclusions and carriers driving in the opposite direction. "You can't have the brokers saying, 'Oh, we're not going to tolerate the standalone market,' and Chubb, for example, saying, 'This is going to be a standalone market,' simultaneously."
The Lloyd's Market Association has already drafted and filed AI endorsements and exclusions - a process that took 20 years in cyber but has been completed within two years for AI. That pace tells its own story about where the market is heading.
Carriers are rapidly introducing sweeping AI exclusions into renewals for cyber, E&O, and general liability. Brokers who have not explicitly addressed AI in a client's program review ahead of their next renewal cycle risk leaving them with uncovered claims - and their own E&O exposure in the process.
Read more: Your clients think cyber covers their AI risk. Most of the time, it doesn't
Testudo distributes through the wholesale channel, positioning AI liability as an E&S product - the natural home, Lewin-Smith argues, for a risk that is new, evolving, and poorly suited to standard lines. Its policy is designed to respond to third-party claims arising from AI-generated outputs, covering legal costs and damages where most carriers are still silent on the exposure and that silence carries a real cost for any broker who hasn't reviewed their client's program for the new ISO exclusion language. The Lloyd's-backed program, expanded with Atrium and QBE capacity in March 2026, reaches $9.25 million per insured.
The underwriting approach is built around proprietary litigation data rather than invasive audits or technical integrations - a deliberate design choice for a line where most commercial buyers are unwilling to allow third-party access to AI systems that may contain sensitive data or non-public information.
Lewin-Smith's read on the broader market is that a large uninsured AI loss - when it comes - will force the conversation that many buyers are still deferring. "The claims are absolutely happening in the market," he said. "The question is at what frequency and severity is it really going to force the market to wake up."
The ISO exclusion forms took effect in January 2026. Brokers whose clients renew in the next six months are already in the window. The ones who check the policy language now, before a claim lands, are the ones who won't be explaining to a client why their coverage failed.
