Guides
AI liability insurance: how insurers are treating artificial intelligence risk
Insurers are adding generative AI exclusions to CGL, D&O and E&O forms while a few markets sell affirmative AI cover. What to check in your policies in 2026.
Insurers are moving on AI in two directions at once. Many are adding exclusions that remove generative AI from traditional liability policies, and a small group of specialty markets sells affirmative AI cover for model errors and related claims. If your business uses AI tools in its work, check your general liability, professional liability, D&O and cyber forms for AI language at this renewal. The exposure was often covered by silence, and that silence is ending.
Why this is a coverage question now
Most liability policies were written without AI in mind. A claim that an AI tool defamed a person, copied protected work, gave a customer wrong information or screened out job applicants unfairly could land on several policies, and carriers could argue about whether any of them responds. Insurers call this "silent AI," in the way "silent cyber" was once debated.
Claims are arriving. Cornerstone Research's 2025 review counted 16 AI-related securities class action filings in 2025, the largest of its trend categories. The SEC settled charges in March 2024 against two investment advisers, Delphia and Global Predictions, for false statements about their use of AI, with civil penalties of $225,000 and $175,000 (SEC release 2024-36). In February 2024 a British Columbia tribunal held Air Canada responsible for wrong fare information its website chatbot gave a customer, in Moffatt v. Air Canada (2024 BCCRT 149), awarding about CA$812. In May 2025 a federal judge in California conditionally certified a nationwide age-discrimination collective action against Workday over its AI applicant-screening tools, per Holland & Knight; there has been no finding of discrimination. Insurance Journal reported in August 2026 on a Gallagher study that AI-related lawsuits rose 978% between 2021 and 2025.
The exclusion wave
ISO's generative AI endorsements. For policies written on the ISO commercial general liability program, Verisk released three optional endorsements in its January 2026 edition, as described by Big "I" Virtual University and Insurance Journal:
| Form | What it removes |
|---|---|
CG 40 47 | Coverage A and B: bodily injury, property damage, and personal and advertising injury arising out of generative AI |
CG 40 48 | Coverage B only: personal and advertising injury arising out of generative AI |
CG 35 08 | Bodily injury and property damage arising out of generative AI under the products/completed operations coverage part |
The forms define generative AI as a machine-based learning system or model trained on data that can create content or responses, such as text, images, audio, video or code. They are optional: each carrier decides whether to attach them. Insurance Business reported on September 16, 2026 that 2,369 generative AI exclusion records were in force across 49 states, and that W. R. Berkley, Chubb, Travelers, Berkshire Hathaway and AIG had each filed to adopt the ISO forms or proprietary equivalents by April. The same article notes that the wording keys on a connection to AI rather than the share of the work AI performed. Policy language is the authority; read the endorsement schedule.
Professional and management liability. Insurance Journal and Insurance Business both report that Berkley introduced an "absolute" AI exclusion for D&O, E&O and fiduciary liability products, broader than the ISO forms. Insurance Journal says it reaches AI use or development, AI-generated content and breaches of AI-related duties. An attorney quoted there expects rapid adoption of some version of an AI exclusion in most lines, with some carriers instead underwriting AI risk for additional premium.
What affirmative AI cover looks like
A few markets have launched cover that names AI performance failures instead of excluding them:
- Armilla, a Lloyd's coverholder, launched an AI liability policy in 2025, with Chaucer among its underwriters, covering legal costs and liabilities from failures such as malfunctions, hallucinations and critical inaccuracies, per Reinsurance News.
- Munich Re's aiSure describes cover for AI performance, discrimination, IP infringement, hallucination, regulatory fines and financial loss, aimed at AI providers who give performance warranties and companies that deploy AI at scale (Munich Re).
- Mayflower Specialty and Hadron announced a program in June 2026 combining E&O, EPL and D&O with a difference-in-conditions and excess layer, said to be the first dedicated affirmative AI liability program in the U.S., per Reinsurance News.
These products are new, and Insurance Business found no admitted standalone product that replaces the CGL coverage removed by the exclusions. Terms, limits and claim triggers differ by program, so a broker should read the wording, including how the policy decides that a loss was "caused by AI."
Where AI exposure sits in a typical program
| Policy | AI-related claim it might see | What to check |
|---|---|---|
| General liability | A defamation or copyright claim over AI-generated marketing; injury from an AI-assisted design | CG 40 47, CG 40 48, CG 35 08 or a carrier's own AI exclusion |
| Professional / tech E&O | A client loss from AI-generated advice or code (Technology E&O insurance for software, SaaS and IT firms) | Any AI exclusion; whether your services definition includes AI-assisted work |
| Media liability | Infringement or defamation in content you publish (media liability insurance: defamation, copyright and content claims) | Whether generated content is addressed |
| EPL | An applicant says an AI hiring tool is discriminatory | Whether the policy covers claims about vendor-provided tools |
| D&O | A securities claim over overstated AI capabilities, or a governance suit | AI exclusions and the entity coverage; see Securities class actions and the D&O market |
| Cyber / crime | Deepfake fraud, data leaks into a model | See Deepfakes and AI-driven fraud: what cyber and crime policies cover |
A hypothetical
Say a 40-person staffing firm adopts an AI tool that ranks applicants. A rejected candidate alleges age discrimination, and a client company sues for losses after the tool misrepresented a candidate's credentials. The firm's EPL policy might respond to the first claim, but only if it has no AI exclusion and treats vendor-provided tools as part of the firm's employment practices. The firm's professional liability policy might respond to the client's claim, if it has not been endorsed with an AI exclusion. The vendor contract may offer an indemnity, but only up to a cap. Without checking the three documents in advance, the firm cannot say who pays.
What a business can do now
- Inventory your AI use. List the tools, who uses them, what decisions they inform and which are vendor-supplied.
- Ask your broker for a policy-by-policy AI review, and for the endorsement lists, before renewal. Ask which form numbers appear and whether the carrier intends to add them.
- Compare the options. The choices are to accept the exclusion, negotiate it away or narrow it, or buy affirmative cover. Each carries a cost.
- Document governance. Human review of outputs, approval protocols, testing and logs are what carriers increasingly look for, according to Insurance Business.
- Fix the contracts. Look for indemnities, warranties and liability caps in vendor agreements, and watch what you promise your own customers (contractual risk transfer: shifting liability by contract and insurance).
- Be accurate in public statements about what your AI does, since overstated claims are an SEC and shareholder issue.
- Watch regulation. Insurers face their own AI rules; see AI regulation in insurance: the NAIC bulletin and state rules.
Key takeaways
- Verisk's optional ISO generative AI exclusions,
CG 40 47,CG 40 48andCG 35 08, became available in January 2026, and carriers including Berkley, Chubb, Travelers and AIG have filed to adopt them or equivalents. - Berkley's broader AI exclusion for D&O, E&O and fiduciary shows the trend extending beyond general liability.
- Affirmative AI products exist from Lloyd's-backed and specialty markets but are new, with terms that vary widely.
- AI claims are real: securities filings, a chatbot misstatement case, an AI hiring class action and SEC "AI washing" penalties.
- The practical steps are an AI inventory, a policy-by-policy review before renewal, documented governance and attention to vendor contracts.