AI insurance policy exclusions are clauses written into commercial insurance policies that remove coverage for losses arising from the use, deployment, or output of artificial intelligence systems. As of August 2026, these exclusions have become one of the most contested issues in commercial insurance, with insurers across general liability, professional liability (E&O), cyber, and directors and officers lines adding AI-specific language to policies. Understanding exactly what these exclusions remove, what they leave intact, and how to negotiate around them is now a core competency for any business that builds, sells, or relies on AI systems.
What AI Exclusions Actually Are
Also worth reading: What is acute onset of pre-existing condition coverage in travel insurance and how does it work? · What are loss of use coverage limits for renters insurance, and how much coverage do I actually need? · Does my credit card's primary rental car coverage really replace the rental company's insurance?
An AI exclusion is a policy provision that states, in effect, that the insurer will not pay for claims connected to artificial intelligence. The wording varies by carrier, but most follow a similar structure: they define AI broadly (often including machine learning models, large language models, generative AI outputs, and automated decision systems), then exclude bodily injury, property damage, personal injury, or professional losses 'arising out of, based upon, or attributable to' the use of that AI. Some exclusions are absolute; others are carve-backs that restore coverage if the AI was used under specific conditions, such as with human review of outputs or within a certified governance framework.
The reason insurers insert these clauses is loss uncertainty. Traditional actuarial models rely on decades of claims data, and AI-related losses simply do not have that history. A generative model that produces defamatory content, a hiring algorithm that discriminates, or an autonomous system that causes physical damage can generate losses at a scale and frequency that underwriters cannot yet price. Rather than absorb that unquantified risk silently, carriers have increasingly chosen to name it and exclude it. Industry reporting through 2025 and into 2026 shows a marked acceleration: what began as occasional endorsements on technology E&O policies has spread into standard commercial general liability forms for businesses of all kinds, not just software companies.
Why Insurers Are Adding AI Exclusions Now
The timing is not accidental. Three forces converged. First, adoption exploded: surveys throughout 2024 and 2025 showed that a majority of mid-sized and large enterprises had deployed generative AI in some operational capacity, meaning the exposure base grew faster than loss data could accumulate. Second, early litigation emerged. Claims involving algorithmic bias in hiring, AI-generated content infringing copyright, chatbots giving negligent advice, and automated underwriting errors began appearing in courts, giving insurers concrete examples of loss scenarios they had not priced. Third, regulatory risk sharpened. The EU AI Act's phased obligations, state-level AI statutes in the United States, and sector regulators' scrutiny of automated decision-making created a new category of fines and defense costs that insurers were reluctant to cover by default.
The result, as reported by Bloomberg Law and Claims Journal, is that policyholder advocates have raised alarms about coverage gaps opening faster than businesses realize. A company that bought a general liability policy in 2023 may renew in 2026 and find, buried in an endorsement schedule, that its most significant emerging risk, its AI stack, is now excluded. Many policyholders do not discover the gap until a claim is denied, which is the worst possible moment.
The Main Types of AI Exclusions
AI exclusions are not monolithic. Knowing the taxonomy helps you read your own policy intelligently. The most common categories as of mid-2026 include:
Absolute AI exclusions, which bar coverage for any claim arising from AI use, full stop. These are the harshest form and are typically found in standard GL policies issued by carriers with no appetite for AI risk. Carve-back exclusions, which exclude AI losses except where defined safeguards were in place, such as human-in-the-loop review, documented model validation, or use of an approved vendor list. Silent exclusions, where the policy does not mention AI at all but existing exclusions for professional services, data, or 'electronic means' are interpreted by the insurer to sweep in AI claims. Regulatory exclusions, which remove coverage for fines, penalties, and defense costs arising from AI-specific statutes like the EU AI Act or state automated decision-making laws. Intellectual property exclusions, which target the fastest-growing AI claim type: allegations that AI outputs infringe copyright, trademark, or trade secrets.
Each type carries different consequences. An absolute exclusion means an AI-related lawsuit is entirely uninsured unless you buy back coverage elsewhere. A carve-back exclusion rewards governance: companies that can document human oversight and model testing may retain coverage that competitors lose.
Comparison: Standard GL Policy vs. AI-Specific Coverage
The practical question for most businesses is how a traditional policy with AI exclusions compares against dedicated AI coverage products that have entered the market. The table below summarizes the differences as they stand in 2026.
| Feature | Standard GL/E&O with AI Exclusion | Dedicated AI Liability Policy |
|---|---|---|
| AI-related bodily injury/property damage | Excluded or heavily conditioned | Covered, subject to underwriting review |
| Algorithmic bias / discrimination claims | Typically excluded | Often covered, sometimes with sub-limits |
| IP infringement from AI outputs | Usually excluded | Covered on some forms; sub-limits of $250K–$5M common |
| Regulatory fines (EU AI Act, state laws) | Excluded where insurable by law | Limited coverage; insurability varies by jurisdiction |
| Premium impact | Lower base premium, hidden gap | Typically 15–40% higher than comparable E&O |
| Underwriting requirements | Minimal | Documentation of AI governance, model testing, human oversight |
| Claim certainty | High denial risk for AI claims | Clearer coverage intent, still subject to exclusions |
How to Audit Your Current Policies for AI Gaps
A structured audit takes most organizations two to four weeks and should happen before renewal, not after a claim. Start by pulling every active policy: general liability, professional liability, cyber, D&O, product liability, and any media liability coverage. Search each policy and its endorsements for the terms 'artificial intelligence,' 'machine learning,' 'algorithm,' 'automated,' 'generative,' and 'model.' Note that exclusions often live in endorsements with innocuous names, so read the full endorsement schedule rather than relying on the base form.
Next, map your actual AI exposure against what you find. Inventory every AI system your business uses or sells, including embedded third-party AI in software you resell, marketing tools that generate content, and customer-facing chatbots. For each, ask which policy would respond to a claim and check whether the AI exclusion kills that response. Pay particular attention to the definition of AI in the exclusion: some definitions are so broad they capture ordinary automation and even spreadsheet macros, which means the exclusion may be far wider than you assumed.
Finally, quantify the gap. If your largest plausible AI claim, say a discrimination class action arising from an automated hiring tool, would generate $3 million in defense costs and your only relevant coverage is excluded, that is a $3 million uninsured exposure. Present that number to your broker and carriers at renewal with a request for either removal of the exclusion, a carve-back, or a quote for supplemental coverage.
Negotiation Strategies That Work
Carriers do add AI exclusions by default, but they frequently negotiate, especially for insureds with documented governance. The single most effective lever is evidence of AI risk management: a written AI use policy, human review requirements for high-stakes outputs, model documentation, bias testing results, vendor due diligence files, and incident response procedures for AI failures. Underwriters in 2026 increasingly price governance directly; insureds who present a mature framework have reported exclusion removals or favorable carve-backs that uninsured-risk peers did not receive.
Second, negotiate the definition of AI itself. A definition limited to 'autonomous machine learning systems' is far narrower than one covering 'any algorithmic or automated process.' Narrowing the definition can shrink the exclusion dramatically without the carrier conceding the underlying point. Third, seek carve-backs tied to specific conditions you already meet, such as coverage restored where outputs receive qualified human review before external release. Fourth, consider a separate AI liability policy or endorsement to sit alongside the excluded policy, so the gap is filled explicitly rather than through contested interpretation. Law firm analyses, including guidance published by Honigman, emphasize that these negotiations should also flow into technology contracts: indemnification clauses with AI vendors and customers can shift or share losses that insurance no longer covers.
Common Mistakes Businesses Make
The most expensive mistake is assuming your existing policy covers AI because it covers 'technology' generally. Courts and carriers read exclusions literally, and a claim denial letter citing an AI endorsement is not the moment to learn the difference. The second mistake is relying on your AI vendor's insurance. Vendor policies rarely extend to your liability for deploying their tools, and vendor contracts frequently cap indemnification at fees paid, which may be a fraction of your loss. Third, businesses often ignore the interaction between exclusions: an AI exclusion in the GL policy combined with a professional services exclusion in the cyber policy can leave a claim with no home at all, a gap that only appears when you read the policies together.
Fourth, some organizations overcorrect and buy the first AI policy offered, accepting sub-limits and exclusions that a competitive market would improve. The AI liability market has grown rapidly, with multiple carriers and MGAs now competing, and pricing for similar coverage has varied by 30% or more between quotes. Fifth, companies treat the audit as a one-time exercise. Exclusion language is evolving quarterly; a policy reviewed in early 2025 may be materially different at a 2026 renewal.
When to Act and What It Costs
Act at renewal, and start 90 days before it. Renewal is when carriers are most willing to negotiate terms, and a broker armed with an AI exposure inventory and governance documentation can request exclusion removals, carve-backs, or supplemental quotes as part of the submission. Waiting until mid-term limits your options to endorsements carriers may decline to issue.
On cost: dedicated AI liability coverage in 2026 typically prices between $5,000 and $50,000 annually for mid-market companies with $1 million to $5 million limits, depending on industry, AI use intensity, and governance maturity. High-risk uses, such as autonomous physical systems or AI in regulated financial decisions, can push premiums higher, sometimes 40% above comparable technology E&O. Carve-back negotiations on existing policies are usually cheaper than standalone policies but may come with reduced limits or higher retentions. Weigh these costs against the exposure: a single uncovered algorithmic discrimination class action routinely exceeds $1 million in defense costs alone, before any settlement.
The Bottom Line
AI exclusions are now a standard feature of the commercial insurance market, not an anomaly, and the trend reported by Claims Journal, Bloomberg Law, and Insurance Journal through 2026 points toward wider adoption, not narrower. Businesses that treat this as a paperwork issue will discover uninsured losses at claim time. Businesses that inventory their AI exposure, audit their policies, document governance, and negotiate with carriers, ideally with a broker who specializes in AI risk, can close most of the gap at a defensible cost. The exclusions are not going away; the question is whether your coverage strategy acknowledges them before your first AI-related claim does.