Carriers are quietly rewriting professional liability policies to exclude generative AI. Architecture and engineering firms still covered today may not be at the next renewal. Here is what design professionals need to know before signing the next E&O policy.
Architecture and engineering firms are using generative AI faster than insurance carriers can rewrite their policies. That gap is closing in twenty twenty-six. And it is closing in a way that could leave design firms exposed at the worst possible moment, right at policy renewal. Here is the situation. Most professional liability policies in force today still cover AI-assisted design work. The standard errors and omissions policy responds to the negligent act, not the tool that helped produce it. Whether the error came from ChatGPT, a Midjourney rendering, or an AI plugin inside a CAD platform, the coverage analysis has been the same as for any other professional negligence claim. That standard answer is breaking down. Three things shifted at the start of twenty twenty-six. First, Verisk released two new exclusion forms effective January first, twenty twenty-six. The forms give carriers ready-made language to exclude losses arising from generative AI. They apply to commercial general liability first. But the industry signal is directional. Some E and O forms are being rewritten to match. Second, Berkley introduced what the market is calling an absolute AI exclusion. The endorsement applies to directors and officers, errors and omissions, and fiduciary liability products. It names ChatGPT, Bard, Midjourney, and DALL-E specifically. Hamilton followed with its own generative AI exclusion. Philadelphia Insurance and Hamilton Select went further, excluding AI-related claims from E and O entirely. Third, AIG, Great American, and W. R. Berkley have filed for regulatory approval to introduce their own versions. Those filings do not show up overnight on every policy. They surface at renewal, in endorsements that many firms skim past. So here is the practical question every design firm needs to answer this year. Does the current policy actually contain AI exclusion language? And if it does not yet, will the renewal change that? A four-step audit answers both questions in about thirty minutes. Step one. Check the declarations page for endorsement schedules. Any form number that starts with C G four zero is a Verisk general liability exclusion. Carrier-specific endorsements that mention artificial intelligence, generative AI, or machine learning warrant a closer read. Step two. Search the policy body for the word artificial. A policy without that word is usually silent on AI, which means coverage follows the standard negligent-act trigger. Step three. Read the definitions section. Some carriers do not add a new exclusion at all. They narrow the definition of professional services or covered products to carve out work produced by non-human systems. The carve-out may never appear in the exclusions section. Step four. Confirm the retroactive date. Claims-made coverage often preserves protection for past AI-assisted work even if the renewal contains a new exclusion. The exposure here is real. Stanford Law School research documented hallucination rates between fifty-eight and eighty-eight percent on general-purpose AI tools. Specialized so-called legal-grade systems still show hallucination rates between twenty and thirty-three percent. For a design firm, a hallucinated beam specification or a nonexistent building code citation is not an abstract risk. It is a claim waiting to happen. Finding an AI exclusion is not a crisis. It is a negotiation starting point. Carriers have shown a willingness to walk back the standard-form language for firms that demonstrate active AI risk management. A two-page internal policy covering tool approval, verification steps, client disclosure, and licensed professional sign-off gives a broker something concrete to negotiate. A carve-back for supervised AI use is the most common negotiated outcome. The bottom line for twenty-twenty-six and twenty-twenty-seven renewals. Firms caught off guard are usually the ones who assumed E and O would roll over with identical language. That assumption needs to be verified every single year now. AI exclusion adoption varies meaningfully across carriers, so an independent broker with access to multiple markets can place coverage that a direct writer cannot. For design firms that want a complete emerging-risk analysis, Risk Specialty Group offers a Full three sixty review covering professional liability, cyber, and the AI-specific exposures that increasingly sit between them. The team is reachable at risk-specialty-group.com.
Risk Specialty Group
City: Houston
Address: 675 Bering Dr.
Website: https://riskspecialtygroup.com/