Who's Liable When AI Misses a Construction Drawing Error?

Does AI shift liability away from the licensed professional?

AI does not shift liability away from the licensed professional, because licensure and the duty of care stay with the architect, engineer, or reviewer of record regardless of what software assisted the review. AI is a tool, not a licensed party, so it cannot absorb professional liability the way a delegated engineer or subconsultant can. A firm that lets an AI flag substitute for a licensed sign-off has a process problem, not a liability shield.

What do current contracts actually say about AI?

Most current construction contracts say nothing about AI, because standard-form agreements like AIA A201 and ConsensusDocs predate the current wave of AI-assisted review and don't address it directly. Browne Jacobson's 2026 horizon-scanning analysis describes this as a real contractual gap: when an AI system contributes to a defect, delay, or safety incident, determining liability between the contractor, the AI vendor, and other parties gets complicated fast, and bespoke contract language covering disclosure, human oversight, and liability allocation is increasingly necessary rather than optional.

Who bears the risk when an owner requires AI review?

An emerging principle in legal commentary, not settled case law, holds that liability allocation should track who mandated AI use. If a project owner requires AI-assisted design or review, the reasoning goes, the owner should bear the cost of remedying an AI-related error, since the owner directed the process. Construction Law Insights' February 2026 analysis walks through this allocation-by-usage principle in more depth, alongside the broader risk picture for AI adoption in real estate and construction.

How is insurance responding?

Some professional liability carriers have already moved to exclude AI claims from certain policy lines. Philadelphia Insurance and Hamilton Select have excluded AI claims from certain professional liability lines, which directly affects what coverage is available to firms relying on AI tools for design or review work. Don't assume your existing E&O policy covers AI-assisted findings just because it covers your firm's general professional work. Ask your broker in writing, specifically about AI-assisted workflows, before you build a process around one.

What changed in 2026 for AI liability and insurance?

The biggest change is on the insurance side. On January 1, 2026, Verisk's CG 40 47 and CG 40 48 endorsements took effect, giving commercial general liability carriers a standard way to exclude generative-AI claims. Dan Cumberland Labs' summary also points to Berkley's Form PC 51380, an exclusion for claims arising from AI use "by any person or entity" across D&O, E&O, and fiduciary lines. Wording like that can reach AI embedded in a vendor's software even when your firm never switched it on, so read the exclusion language itself, not a summary of it.

The contract forms have not caught up. Commentators report that the 2026 editions of AIA and EJCDC documents still contain no AI-specific clauses, with AIA pointing firms to existing approved language rather than writing new terms. ConsensusDocs has taken the other route and tells parties to negotiate AI provisions directly, covering human verification, error allocation, indemnity, and flow-down to subcontractors. The practical result is that the four written items below matter more in 2026 than they did a year ago.

One more shift: the AI tools themselves now sit inside the platforms you already use. Procore AI's Drawing Analyst and Autodesk's drawing features are chosen, and often paid for, inside the project platform, which makes "who selected the tool" harder to answer after the fact. If an agent runs inside your document system, record who enabled it and what it was configured to check.

What should be in writing before you rely on AI drawing review?

Before you rely on AI drawing review, put four items in writing: AI's screening role, a named human sign-off, who selected and paid for the tool, and written insurance confirmation.

1. Define AI's role explicitly. State that it screens and flags; it does not replace licensed review or QA/QC sign-off.
2. Require a documented human sign-off on every AI-flagged issue, whether it was acted on or dismissed, with the reviewer's name attached.
3. Name who selected and paid for the tool. If the owner mandated it, put that in the contract, since it affects the allocation-by-usage principle above.
4. Confirm insurance coverage in writing. Get your broker to state explicitly that your E&O and general liability policies extend to AI-assisted workflows.

How does Helonic approach AI review liability with customers?

Helonic has never been marketed as a replacement for licensed review. Every finding is tied to an exact page location and, where applicable, the governing code compliance section, so a qualified reviewer can confirm or dismiss it in seconds with a documented decision, not a rubber stamp. That accept/reject audit trail is what a contract needs to point to when the liability question comes up, and it's a theme we cover in construction AI myths and where AI plan review is and isn't reliable.

For teams building contract language around AI use, our construction lawyers solutions page covers how drawing ambiguities and coordination gaps surface in disputes and claims, and how drawing issues contribute to insurance claims and litigation more broadly. If your team is drafting AI-specific contract language now, see how proactive drawing review reduces litigation exposure for the underlying pattern.

See the fuller picture of what AI can and can't be trusted for in our AI for construction drawings pillar.

AI Drawing Review Liability: FAQ