Short answer
AI plan review liability depends on who relied on the output and how the work was delivered. Most tools describe their results as advisory, but design firms and building departments still act on them. A design firm remains responsible for its own design, and the software company may face its own claim if a client says the tool failed. Policy wording, contracts and underwriting decide what coverage applies.
Advisory output still gets relied on
Most AI code checkers state in their terms that output is advisory and that users must verify results. That language sets expectations, and it can support a defense.
It does not stop a client from bringing a claim. If a design firm submits drawings after the tool reported no issues, and a problem surfaces later, the allegation can still focus on the tool’s output. Whether a claim succeeds depends on the facts, the contract and the applicable law, and qualified counsel can answer those legal questions.
Missed issues vs. false flags
A tool can fail in two directions, and both cost money:
- Missed issues. The tool does not flag a real code problem, which can lead to rework, permit rejection or construction delay.
- False flags. The tool reports problems that do not exist, which can trigger unnecessary redesign and cause a firm to miss a permit window.
Most people think about missed issues first. A client who loses a permit window because of a wrong flag can also claim a financial loss.
Local amendments are the hidden risk in AI plan review liability
A wrong code edition or a missed local amendment creates a different kind of problem. A single error can repeat across every project reviewed in that jurisdiction.
That pattern matters to carriers because one underlying mistake can produce many claims. Software companies should know how their policy treats related claims, and whether limits apply once for the whole group or separately for each project. Policy wording varies, so ask your broker to confirm how yours reads.
When a human reviewer signs off
Some companies add a licensed reviewer who confirms the tool’s output before it reaches the client. That step improves quality, and it changes how a carrier sees the work.
A licensed person confirming code compliance looks like professional plan review. Many technology E&O forms exclude work performed in the capacity of an architect or engineer, so the policy may not respond to that service. Our article on [technology E&O vs professional liability for AEC software companies] explains where that line often falls.
Selling to building departments
Public agencies often attach their own insurance and indemnity terms to a contract. These can include:
- Minimum E&O, cyber and general liability limits
- Specific indemnity wording that goes beyond ordinary contract terms
- Requirements to name the agency on certain policies
Review these terms before you sign, and confirm your policy fits each one.
For design firms using these tools
A design firm’s professional liability policy still carries its design work. Using an AI checker does not shift responsibility to the software vendor. If the tool misses something and the firm submits the drawings, the firm remains responsible for its own design.
Any claim against the vendor is a separate question, and it depends on the vendor’s contract. Tell your broker when your firm relies on these tools, since carriers may ask about quality control procedures.
Example: the missed local amendment
A software company sells an AI checker to design firms in several cities. The tool uses a base code edition and misses a local amendment on egress in one city. Three firms submit drawings after the tool reports no issues. The city rejects all three, and each firm pays for redesign and loses weeks on its schedule.
The firms allege the tool’s output was wrong. The company points to its advisory language, which supports its defense, but it still has to respond to the allegations. Whether the claims are treated as one or three under the policy depends on the wording. This example is hypothetical, and coverage depends on the policy terms, exclusions and application.
What to check
- Does your terms of service match how clients actually use the tool?
- Does your policy treat the output as a technology service, a professional service or both?
- Does your form exclude work performed as an architect or engineer when a licensed reviewer is involved?
- How does your policy handle related claims across many projects?
- Do public contracts require limits or indemnity wording your policy does not meet?
- Does your design firm’s professional liability carrier know you use AI review tools?
Related questions
Is AI plan review a technology or professional service?
It can be both. A purely automated check may read as technology work, while human sign-off often reads as professional services.
If we use an AI checker, can we blame the vendor?
You remain responsible for your own design. Any claim against a vendor is a separate matter that depends on its contract and the facts.
What do building departments require?
Public contracts often require E&O, cyber and general liability coverage, along with specific indemnity wording. Requirements differ by agency, so review each contract.