What is an AI insurance policy exclusions checklist and how do I know if my business coverage excludes AI-related claims?

Artificial intelligence has moved from a novelty to a core business function in a remarkably short period, and insurers have responded in ways that most policyholders have not yet noticed. According to the Center for Strategic and International Studies, the insurance industry has been retreating from AI-related risks through a quiet but accelerating wave of exclusions, endorsements, and coverage carve-outs. This retreat threatens to slow AI adoption across entire sectors, because companies that believed they were covered for AI-related errors, discrimination, or system failures are discovering gaps only when a claim is denied. An AI insurance policy exclusions checklist is the systematic review process that identifies these gaps before they become financial losses. It involves reading every policy in your program, general liability, professional liability, cyber, directors and officers, and property, and asking one question of each: if an AI system causes harm, does this policy respond, exclude, or stay silent? This guide walks through the full checklist process, explains where AI exclusions hide, and gives you the practical steps to close coverage gaps before your renewal date.

Why AI Exclusions Are Spreading Through the Insurance Market

Also worth reading: Do AI Agent Insurance Exclusions Leave Businesses Uncovered in 2026? · AI Insurance Exclusions in 2026: What They Cover and What They Do Not? · What Should an AI Insurance Audit Checklist Include for 2025?

The insurance market's reaction to AI follows a familiar pattern that played out with cyber liability between 2015 and 2020. Insurers initially absorbed AI-related claims under existing policy language because nobody wrote specific exclusions. As loss patterns emerged, model errors, hallucinated outputs, algorithmic discrimination claims, AI agents taking unauthorized actions, carriers began drafting language to push these risks back to insureds. Dentons and Jones Day have both documented a wave of AI exclusions appearing in directors and officers (D&O) and errors and omissions (E&O) programs, while Risk & Insurance has reported that general liability policies increasingly contain AI carve-outs buried in endorsements rather than in the main exclusion section.

The scale of this shift matters. Industry reporting through 2025 and 2026 indicates that a growing share of professional liability renewals now include some form of AI-specific exclusion language, and Dark Reading reported that while cyber insurance rates were dropping through this period, the exclusions within those cheaper policies were widening at the same time. In other words, some businesses paid less for coverage that quietly covered less. The practical consequence is that two companies with identical operations can hold policies worth dramatically different amounts depending on which endorsements their broker negotiated. This is precisely why a written checklist matters: exclusions are not announced, they are inserted, and the only defense is deliberate review.

The Core AI Insurance Policy Exclusions Checklist

A usable checklist has eight review points, and each one should be applied to every policy in your program. First, inventory every place AI touches your business: products you sell with embedded AI, services you deliver using AI tools, internal decisions AI influences such as hiring or underwriting, and autonomous agents that act on your behalf. Second, locate every exclusion or endorsement in your policies that mentions artificial intelligence, machine learning, algorithms, automated decision-making, or generative AI. The language varies enormously between carriers, so keyword search the full policy document rather than relying on the summary. Third, check whether your professional liability policy treats AI-assisted work as your professional service or as something excluded. Fourth, examine your general liability policy for bodily injury, property damage, and personal injury carve-outs connected to AI systems. Fifth, review your cyber policy for whether AI agent actions, prompt injection losses, or deepfake-enabled fraud fall within or outside coverage. Sixth, confirm whether D&O coverage responds to shareholder or regulatory claims arising from AI governance failures. Seventh, verify how each policy defines the AI activity being excluded, because overly broad definitions can sweep in ordinary software. Eighth, document every gap in writing and assign a remediation owner with a deadline tied to your renewal date.

The reason this checklist works is that AI exclusions fail in predictable places. Most policyholders who suffer a denied AI claim discover the loss fell between two policies: the professional liability carrier said the loss was a technology failure, and the cyber carrier said it was a professional error. Your checklist exists to force that conversation to happen before the claim, not during it.

Where AI Exclusions Hide: Policy by Policy

Each line of coverage hides AI exclusions differently, and knowing the typical locations saves hours of review time. In general liability policies, AI exclusions frequently appear as amendments to the standard CGL exclusion list or as separate endorsements, sometimes described in industry press as the clause already sitting on contractor clients' policies without their knowledge. These endorsements commonly exclude bodily injury or property damage arising from autonomous or semi-autonomous systems, which matters enormously for construction firms using AI-guided equipment or manufacturers shipping products with embedded machine learning.

Professional liability and E&O policies tend to exclude claims arising from AI-generated work product unless a specific endorsement buys the coverage back. Dentons has documented insurers adding exclusions for claims arising from the use of artificial intelligence in rendering professional services, which would otherwise be the core of an architect, engineer, consultant, or marketing agency's work. Cyber policies present the most active frontier: Reuters reported that as AI agents have gone rogue, performing unauthorized transactions or exfiltrating data, cyber insurers have scrambled to adapt, and Insurance Business covered cases where OpenAI-style autonomous agent failures exposed genuine gaps in cyber coverage. Finally, D&O policies increasingly exclude claims tied to AI governance failures, algorithmic discrimination lawsuits, or misstatements about a company's AI capabilities, the latter becoming a securities fraud risk as regulators scrutinize AI washing.

Comparison: Insuring AI Risk Through Endorsements Versus Standalone AI Policies

Once your checklist identifies gaps, you generally have three remediation paths: negotiate endorsements onto existing policies, buy standalone AI liability coverage, or accept the risk. The endorsement and standalone routes differ enough that the choice deserves deliberate analysis rather than default.

Neither option is automatically superior. Endorsements are cheaper and faster but often carry restrictive definitions, such as excluding claims arising from training data problems even when the endorsement covers model output errors. Standalone policies offer cleaner coverage but introduce the inter-policy gap problem again, since a standalone AI form may still exclude what your cyber policy previously covered. A thorough checklist review treats both options as negotiable and prices them against the actual exposure quantified in your AI inventory.

Common Mistakes Businesses Make When Reviewing AI Exclusions

The most common and expensive mistake is assuming that silence means coverage. If your policy does not mention AI at all, that is not reassurance; it is ambiguity, and ambiguity gets resolved through litigation at your expense. A second mistake is reviewing only the exclusions section of the policy. Endorsements and definitions modify the base form, and in many 2025 and 2026 policyholders' programs the AI restriction lives entirely in an endorsement numbered and paginated away from the main exclusions. Third, businesses routinely forget that their vendor relationships do not transfer coverage. If a third-party AI tool causes your loss, the vendor's insurance rarely responds to your liability, and indemnification clauses in vendor contracts are only as good as the vendor's balance sheet.

A fourth mistake is treating the checklist as a one-time exercise. Policy language changes at every renewal, and carriers have been adding AI exclusions mid-cycle in some cases via endorsement. A fifth error is overlooking occupational and workplace dimensions: research on AI-related hazards notes that many AI harms are psychosocial in nature, such as algorithmic management stress or AI-influenced termination decisions, which can generate employment practices liability claims that neither GL nor cyber policies address. Finally, many companies over-rely on their broker's summary rather than reading the actual policy text. Brokers are valuable, but carriers have been inserting AI language faster than broker training materials have kept up, and the person with legal responsibility for the coverage gap is the insured, not the intermediary.

When to Act: Timing Your Checklist Against Renewal Cycles

The single best moment to run an AI exclusions checklist is 90 to 120 days before your renewal date. This window gives your broker time to shop the AI exposure to alternative carriers, request coverage buy-back endorsements, and negotiate definition changes. Carriers have the most flexibility during underwriting; after a policy binds, endorsements typically require mid-term endorsements that carriers resist granting without a corresponding premium adjustment or, more commonly, refuse entirely. If your renewal falls in the first quarter, meaning January through March, the window to act is effectively now, because underwriters finalize appetite guidance in the preceding quarter.

Certain trigger events should force an off-cycle review regardless of renewal timing. These include deploying autonomous AI agents that take actions without human review, launching a customer-facing AI product, using AI in employment or lending decisions where regulatory exposure exists, receiving notice that your carrier is adding AI language at renewal, or learning that a peer in your industry suffered an AI-related claim. Companies that wait for a claim to discover an exclusion have no negotiation leverage at all, since coverage disputes are settled by policy text, not by fairness. The asymmetry is stark: a few hours of checklist work in advance costs almost nothing, while a denied seven-figure claim that traced to a two-page endorsement costs everything.

Cost Considerations and What Closing the Gap Actually Costs

Budgeting for AI risk remediation has three components: the checklist itself, the coverage changes, and the operational controls that make coverage affordable. The checklist costs little if performed internally with policy documents and a structured template, or several thousand dollars if performed by coverage counsel, which is advisable for companies with revenue above roughly $10 million or complex AI deployments. Coverage changes vary widely. Buying back AI coverage through endorsements on E&O or GL policies has typically added low single-digit to mid-teens percentage increases to the affected line's premium. Standalone AI liability policies for mid-market companies frequently price between $5,000 and $50,000 annually depending on revenue, AI use case, and governance maturity, while larger enterprises with substantial AI product lines can see materially higher figures.

Insurers reward underwriting discipline. Companies that can document AI governance, human oversight of model outputs, testing protocols, and vendor management, routinely receive better terms than companies that cannot, and some standalone AI markets will decline to quote businesses lacking basic governance documentation. This means the cheapest coverage improvement is often not an insurance purchase at all but an internal control that converts an uninsurable risk into an insurable one. Treat the checklist not as a compliance chore but as the mechanism that prices your actual AI exposure accurately for the first time. Businesses using an AI Insurance Checker tool can accelerate this process by scanning policy documents for exclusion language and flagging the specific endorsements that require human review, reducing what was a multi-day manual review into a structured workflow.

The Bottom Line for Business Owners and Risk Managers

The insurance market's retreat from AI is real, documented, and accelerating, and it will not pause for your renewal cycle. Exclusions are appearing across general liability, professional liability, cyber, and D&O lines, often in endorsements that policyholders never read, and the market dynamic of dropping cyber rates paired with widening exclusions means price signals alone will not warn you. Running an AI insurance policy exclusions checklist is the only reliable way to know whether an AI-related claim, an algorithmic discrimination suit, a hallucinated output that damages a client, an autonomous agent's unauthorized transaction, or a misrepresentation claim about your AI capabilities, would be paid or denied. The checklist takes hours, costs little, and must be repeated annually because carriers change language every cycle. Businesses that treat AI coverage review with the same seriousness they now treat cybersecurity will negotiate from strength; those that do not will learn their policy's true terms from a denial letter. Start with your renewal date, work backward 90 days, and read the endorsements.