Two insurance exclusions walked into a construction project at the same time. One stripped AI-related claims from your architect’s errors-and-omissions policy. The other stripped them from your general contractor’s commercial general liability coverage. Both exclusions use language derived from the same ISO endorsement, CG 40 47, released January 1, 2026. Both took effect at the professionals’ most recent policy renewals, and neither professional is required to tell you.
We covered the architect side of this problem three days ago. What we did not calculate then, and what nobody in the insurance press has quantified, is what happens when both exclusions overlap on the same project. The math produces a coverage gap that is wider than either exclusion alone, and it lands squarely on the person who hired both professionals.
Two Policies, One Hole
Commercial general liability insurance and errors-and-omissions insurance protect against different failures. CGL covers bodily injury and property damage arising from operations, while E&O covers financial harm from professional mistakes. On a residential construction project, these two policies form a floor beneath the homeowner: if the architect’s design is wrong, E&O responds; if the builder’s execution is wrong, CGL responds. Between them, most failure modes land somewhere on coverage.
ISO’s CG 40 47 endorsement punches a hole through both.
The endorsement defines “generative artificial intelligence” as “a machine-based learning system or model that is trained on data with the ability to create content or responses, including but not limited to text, images, audio, video, or code.” That definition covers cost estimation software that generates material takeoffs, scheduling tools that produce project timelines, code-compliance platforms that output approval recommendations, and structural analysis programs that generate calculation reports. Every AI tool a modern construction professional touches falls within the language.
Coverage A, bodily injury and property damage: excluded. Coverage B, personal and advertising injury: also excluded. When a carrier adopts CG 40 47 on a CGL policy and a separate carrier adopts comparable language on an E&O policy, the architect and the builder are both uncovered for the same category of error at the same time, on the same project, involving the same homeowner’s money.
Running the Numbers Nobody Ran
ISO forms underpin approximately 82 percent of U.S. property and casualty insurance policies, according to Cohen Seglias Pallas Greenhall & Furman. The American Institute of Architects reports that one-third of architecture firms now use AI in daily operations. Among general contractors, adoption data is sparser, but the ASCE December 2025 survey found 27 percent of AEC professionals using AI, with 94 percent planning to increase adoption in 2026.
Cross-referencing those numbers produces a rough estimate that is nevertheless instructive. If 82 percent of CGL policies will eventually carry the ISO exclusion, and roughly 30 percent of construction professionals are using AI today, approximately one in four residential projects right now involves at least one professional who both uses AI and carries a policy likely to adopt the exclusion at the next renewal cycle. By mid-2027, when the 94 percent planned adoption wave hits and carriers have had two renewal cycles to incorporate exclusion language, the overlap could reach the majority of projects.
That calculation has inputs worth questioning, and the uncertainty runs in both directions. Not every carrier will adopt CG 40 47. Some will see an opportunity to write AI-inclusive coverage at higher premiums. The 82 percent figure represents ISO’s market penetration for form language, not confirmed adoption of this specific endorsement. But the trajectory is one-directional: Berkley, AIG, Philadelphia Insurance, and Hamilton Select have already moved, and Risk Specialty Group reports the adoption pace is accelerating with each renewal cycle.
The International Blind Spot
HCR Law published an analysis on May 12, 2026 that extends the problem beyond U.S. borders. In the United Kingdom, the two dominant standard-form construction contracts, JCT and NEC, contain no provisions addressing AI at all: no allocation of AI-related risk, no disclosure requirements for AI tool usage, and no definition of liability when AI-generated outputs contribute to a defect.
HCR’s conclusion is blunt: “Traditional professional indemnity and construction insurance policies may not respond to AI-related losses.” The U.S. insurance market is at least moving, if only toward exclusion. The UK construction market has not started the conversation. A British homeowner commissioning a build from an architect using AI design tools has no contractual mechanism to allocate the risk of AI failure, and no insurance framework to fall back on when the contract is silent.
That creates two distinct failure modes in two jurisdictions. In the United States, the risk is explicit exclusion: carriers acknowledge the risk exists and refuse to cover it. In the United Kingdom, the risk is contractual silence: nobody has decided who bears the risk, so the answer will be determined by litigation after the damage is done.
What a Homeowner Should Do Before Signing
Ask for certificates of insurance from every licensed professional on your project, and read them carefully. Not the summary page, but the endorsement schedule itself. Look for CG 40 47, CG 40 48, or any language referencing “artificial intelligence,” “machine learning,” or “generative AI” in the exclusions section. Your insurance broker can help you interpret the language if you have one. If you do not, the National Association of Insurance Commissioners maintains a consumer assistance portal that connects homeowners with state-specific resources.
If you find an AI exclusion on either the architect’s E&O policy or the builder’s CGL policy, you have three options, ranked by cost and effectiveness.
First and cheapest: add contract language making the professional responsible for AI-related errors regardless of insurance status. Something like: “Professional’s obligation to perform services in accordance with the applicable standard of care is independent of, and not diminished by, any insurance exclusion related to artificial intelligence, machine learning, or automated tools used in the performance of services.” This does not create insurance where none exists, but it establishes a contractual claim that survives the exclusion, meaning you can pursue the professional’s assets directly if their insurer denies coverage.
Second: require an Owner’s Protective Professional Indemnity (OPPI) policy, a standalone product that covers the owner for losses arising from professional negligence on the project. OPPI policies run $2,000 to $8,000 on a typical residential project, depending on scope and limit. They respond to the homeowner directly, bypassing the question of whether the professional’s own policy covers the claim.
Third and most expensive: hire an independent code reviewer ($500 to $2,000) to audit all AI-touched deliverables before construction begins. This catches errors before they become defects, which is cheaper than litigating coverage after the wall is already wrong.
The Strongest Case Against Worry
Insurance exclusions are not coverage bans. CG 40 47 is an optional endorsement. Carriers that adopt it do so at renewal, one policy at a time, and plenty of carriers have not adopted it yet. Some specialty markets, particularly surplus lines carriers accustomed to writing unusual risks, may develop AI-inclusive products that cover exactly the gap the exclusion creates, at higher premiums but with genuine coverage. The market for AI-specific insurance products is embryonic, not dead.
More importantly, AI tools in construction today are overwhelmingly used as assistants, not autonomous decision-makers. The architect still stamps the drawings, and the contractor still supervises the crew. If a professional uses AI to draft a cost estimate and then reviews it, adjusts it, and signs it, the resulting document is a professional work product, and the AI exclusion may not apply because the loss arose from the professional’s final judgment rather than from the AI output itself, a distinction that courts have not yet tested but that has real legal teeth. But the argument has teeth, and an insurer seeking to deny coverage would need to prove that the AI, rather than the professional’s reviewed and adopted decision, caused the loss.
Limitations of This Analysis
The 25 percent overlap estimate multiplies ISO market penetration (82 percent, from Cohen Seglias) by AI adoption rates (27 to 33 percent, from ASCE and AIA). These are independent data points from different methodologies and different populations, and treating them as directly multiplicable produces an approximation, not a precise measurement. No published data tracks how many carriers have specifically adopted CG 40 47 since its January 2026 release. The OPPI cost range ($2,000 to $8,000) comes from broker-reported figures for residential projects, not from a published rate study. No AI-related construction insurance claim has been litigated to a published decision in any U.S. jurisdiction. UK analysis relies on HCR Law’s May 2026 review; other UK firms may assess the JCT and NEC landscape differently.