Architecture and engineering firms across the country are integrating generative AI into design workflows faster than carriers can rewrite their policies. Most professional liability policies in force today still cover AI-assisted design work as a standard professional negligence matter. The tool that helped produce the deliverable does not change the coverage trigger for the negligent act itself.
That standard answer is breaking down. Carriers are filing new exclusion language with state regulators. Some have already moved. The question facing every design firm in 2026 is no longer whether E&O covers AI-assisted design work in general. It is whether a specific policy, with a specific carrier and a specific renewal date, still answers yes.
Why E&O Coverage for AI Is Shifting in 2026
Three developments converged at the start of 2026 to reshape the market.
Verisk released two endorsement forms (CG 40 47 and CG 40 48) effective January 1, 2026. The forms provide carriers with ready-made language to exclude losses arising from generative AI. They apply to commercial general liability first, but the market signal is directional. Some E&O forms are now being rewritten to match.
Berkley introduced an absolute AI exclusion spanning directors and officers, errors and omissions, and fiduciary liability products. The endorsement 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&O entirely.
AIG, Great American, and W.R. Berkley filed for regulatory approval to introduce their own versions. Filings do not appear overnight on every policy. They surface at renewal, line by line, in endorsements that many firms skim past.
Design firms working with experienced professional liability insurance brokers are getting earlier visibility into carrier-specific changes. Firms working with direct writers often see the new language for the first time at renewal, with a tight signing window.
The Four-Step E&O Audit Every Design Firm Should Run Before Renewal
The audit takes less than thirty minutes with the full policy PDF in hand.
Step one: check the declarations page for endorsement schedules. Look for any form number starting with CG 40, which signals a Verisk general liability exclusion. Look also for carrier-specific AI endorsements, such as Berkley’s PC 51380 or Hamilton’s generative AI exclusion. Any endorsement title that contains “artificial intelligence,” “generative AI,” or “machine learning” warrants a closer read.
Step two: search the policy body for the word “artificial.” Most exclusions appear under a heading that includes “artificial intelligence” or within the exclusions section of the insuring agreement. 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 existing definitions instead. “Professional services” or “covered products” may be redefined to carve out work produced by non-human systems. The carve-out may not appear in the exclusions section because the work was never defined as covered in the first place.
Step four: confirm the retroactive date. Even if the renewal carries an AI exclusion, the claims-made trigger and retroactive date often preserve coverage for past AI-assisted work. A firm that has been using AI tools for two years and gets a new exclusion at the next renewal is not necessarily losing protection for that earlier work.
What AI Use Is Most Likely to Stay Covered
Coverage rarely splits cleanly along AI-versus-no-AI lines. It splits along how the AI was used and what the claim alleges.
AI use that generally remains covered under most current E&O policies includes design work where AI assists a licensed professional who reviews, verifies, and stamps the output. AI-generated renderings used for client visualization typically remain covered when the construction documents themselves are produced by humans. AI-drafted specifications that a licensed engineer reviews, revises, and seals fall in the same category. Internal productivity tools that do not touch the design deliverable, such as scheduling assistants or meeting note summarizers, sit clearly outside the deliverable chain.
AI use that is more likely to trigger a coverage dispute includes deliverables produced primarily by AI and passed through with minimal human review, designs that rely on AI outputs containing hallucinated code references or nonexistent materials, client data uploaded to AI platforms that later surfaces in a breach claim, and work produced by AI tools specifically named in a Berkley, Hamilton, or carrier-specific exclusion endorsement.
The data leakage scenario sits in a particularly gray zone. Client information uploaded to ChatGPT that later surfaces in a breach is generally a cyber liability insurance matter rather than an E&O matter, which is why many design firms now need both lines working together. The broader 2026 picture appears in AI liability insurance for architects coverage analysis.
What to Do When the Audit Surfaces an AI Exclusion
Finding an AI exclusion on a current or renewal policy 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. Many 2026 AI exclusions were added defensively rather than based on actual loss experience. Firms with documented AI governance policies, verification workflows, and licensed professional sign-off procedures often receive broader language than the off-the-shelf endorsement.
A carve-back for supervised AI use is the most common negotiated outcome. The carve-back keeps the exclusion intact for unsupervised or autonomous AI use while restoring coverage for tools used under human oversight. American Institute of Architects Trust guidance on generative AI in architectural practice is worth citing in carrier conversations, since it establishes an industry standard of care for AI use under licensed professional supervision.
Documentation matters as much as the negotiation itself. A two-page internal policy covering tool approval, verification steps, client disclosure, and professional sign-off gives a broker concrete terms to negotiate. It also signals to underwriters that the firm treats AI as a managed risk rather than an open-ended exposure.
A second carrier quote often produces meaningful differences. AI exclusion adoption varies widely. An independent broker with access to multiple markets can place coverage that a direct writer cannot. Risk Specialty Group works with over twenty “A” rated carriers, and the appetite for AI-assisted design work differs significantly among them.
The Bottom Line for 2026 and 2027 Renewals
Firms caught off guard at renewal are usually the ones who assumed E&O would roll over with identical language. In 2026, that assumption needs to be verified annually. The four-step audit takes thirty minutes. The exposure, if it gets skipped, is the entire professional liability tower.
Design firms seeking a comprehensive emerging-risk analysis can request a full 360° review from Risk Specialty Group’s professional liability team.
Content strategy by National Digital Marketing Agency ASTOUNDZ.
Risk Specialty Group
675 Bering Dr.
Suite 175
Houston
TX
77057
United States
