How do we allocate AI liability in customer contracts? — CLO AI Chambers

When it comes up

When the company launches an AI feature, when a large customer asks for AI-specific warranties or indemnities, or when sales escalates AI terms in a deal.

Your options

Broad disclaimer, low cap

Fits when: Early or experimental AI features, low-stakes outputs, and customers who accept beta-style terms.

Watch out: Disclaimers don't override consumer protection or misrepresentation law; large customers will push back.

Defined performance commitments with a standard cap

Fits when: Mature AI features where you can describe and test what the AI does.

Watch out: You must be able to evidence the commitments with testing and monitoring.

AI-specific indemnities (for example, IP in outputs) with conditions

Fits when: Competitive markets where major providers already offer indemnities, and where you have upstream cover.

Watch out: Make sure upstream vendor indemnities match what you give customers, or the gap is yours.

Super-cap or uncapped for specific AI risks

Fits when: Strategic deals with a strong negotiating counterparty, for narrow risks like data misuse.

Watch out: Price it and get finance sign-off; check insurance.

How to decide

Describe the AI feature honestly: what it does, its limits, the customer's review duties.
List the risks: wrong output, IP claims, data misuse, regulatory breach.
Allocate each risk to the party best placed to control it.
Match upstream and downstream: compare what your vendors give you with what you give customers.
Set preferred, fallback and walk-away positions with finance and the business, and record exceptions.

Trade-offs

Generous terms win deals but can leave you carrying risks your vendors won't cover. Tight terms protect the balance sheet but slow sales and may not survive consumer law anyway.

Real examples

In Moffatt v. Air Canada, 2024 BCCRT 149, the tribunal held Air Canada liable for negligent misrepresentation by its website chatbot and rejected the argument that the chatbot was responsible for its own actions. Source: McCarthy Tétrault — Moffatt v. Air Canada: a misrepresentation by an AI chatbot (2024 BCCRT 149) ↗
In September 2024 the FTC announced action against DoNotPay, which marketed a 'robot lawyer', as part of Operation AI Comply; the proposed order included $193,000 in monetary relief and notice to past subscribers about the service's limits. Source: FTC — FTC announces crackdown on deceptive AI claims and schemes (Operation AI Comply, 25 September 2024) ↗

What to check first

What exactly does the AI feature do, and what do sales and marketing say it does?
Which upstream vendor terms (indemnities, caps) back the commitments you're asked to give?
Are customers consumers, where disclaimers have limited effect?
Does your insurance cover AI-related claims?

Learn and use

Sources

Last reviewed 2026-10-01.