Usually, yes. A tribunal has already ruled on it, and it held that a chat window is just part of the website it sits on. The airline that argued its chatbot was a separate legal entity lost, and it paid.
The airline that blamed its own chatbot
In November 2022, Jake Moffatt asked Air Canada’s website chatbot about bereavement fares. The chatbot’s answer said a customer who had already travelled could submit the ticket for a reduced bereavement rate within 90 days of the date the ticket was issued. A separate Air Canada webpage said the bereavement policy did not apply to requests made after travel was completed. Moffatt booked, applied six days later, well inside the 90 days, and was refused.
The dispute went to the Civil Resolution Tribunal in British Columbia. The decision, Moffatt v. Air Canada, 2024 BCCRT 149, carries the date February 14, 2024 and the name of Tribunal Member Christopher C. Rivers.
The airline’s defence is the part worth reading. From paragraph 27 of the decision: “In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission.” The member kept going. “While a chatbot has an interactive component, it is still just a part of Air Canada’s website. It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.”
The airline also argued the correct answer was sitting on another page of its site. Paragraph 28 turned that down too. Air Canada did not explain why the other webpage was inherently more trustworthy than its chatbot, or why customers should have to double-check one part of a website against another part. The tribunal called it negligent misrepresentation.
The order in paragraph 44 was $812.02 in total: $650.88 in damages, $36.14 in pre-judgment interest, and $125 in tribunal fees. Those are numbers you can pull up and check yourself.
Here is the limit, and it is a real one. That is a small claims decision from a tribunal in Canada. It is not US law. No court in your state has to follow it. Read it as a warning about an argument that failed, not as a rule that already applies to your shop.
What US law says about a machine acting for you
The federal E-SIGN Act speaks to this directly. Section 7001(h) of Title 15 of the US Code, read on the official site of the Office of the Law Revision Counsel, where the most recent edition listed for that section is the 2024 Main Edition, dated January 6, 2025, says a contract “may not be denied legal effect, validity, or enforceability solely because its formation, creation, or delivery involved the action of one or more electronic agents so long as the action of any such electronic agent is legally attributable to the person to be bound.”
Section 7006(3) defines an electronic agent as “a computer program or an electronic or other automated means used independently to initiate an action or respond to electronic records or performances in whole or in part without review or action by an individual at the time of the action or response.” A chat widget that answers on its own fits that description.
Now read the tail of that first quote again. It says a deal is not thrown out just for being made by software. It does not say every line your chat types is a binding contract. Whether one specific promise sticks is still an ordinary question of contract law, and that turns on your state and on the facts.
This is not legal advice. Before you act on any of it, ask an attorney licensed in your state or call your state’s consumer protection office.
One state wrote it down in plain words
Utah put the idea in a statute. Title 13, Chapter 77 of the Utah Code covers generative artificial intelligence, consumer disclosures, and enforcement.
Section 13-77-102 reads: “It is not a defense to the violation of any statute administered and enforced by the division under Section 13-2-102 that generative artificial intelligence: (1) made the violative statement; (2) undertook the violative act; or (3) was used in furtherance of the violation.” The copy published by the Utah State Legislature is marked effective May 6, 2026, after an amendment in the 2026 general session.
Section 13-77-105, marked effective May 7, 2025, lets the division director impose an administrative fine of up to $2,500 for each violation of that chapter.
That chapter is Utah’s. If you run a roofing company in Georgia, it is not your statute, and none of this is legal advice. It still shows where this is heading. Your own state’s consumer protection office can tell you which law there covers a wrong price or an invented warranty.
The FTC says there is no AI exemption
On September 25, 2024, the Federal Trade Commission announced five law enforcement actions under the name Operation AI Comply. In the press release, then Chair Lina M. Khan said: “Using AI tools to trick, mislead, or defraud people is illegal. The FTC’s enforcement actions make clear that there is no AI exemption from the laws on the books.” One of the five was DoNotPay, a service sold as a robot lawyer. The FTC announced the final order on February 11, 2025: $193,000 in monetary relief, and notice to the consumers who subscribed between 2021 and 2023.
Be careful how much weight you put on that. Those five cases were about claims companies made about their AI products. None of them was a case about a chatbot telling one customer the wrong thing. What carries over is the principle in the quote, not the cases themselves. Again, that is not legal advice, and an attorney licensed in your state is the one to read your own exposure with you.
Where this bites a trade business
Your chat talks to strangers at 11pm with nobody watching. The places it can commit you are the same places a brand new hire on the phone could commit you:
- A price or a flat rate for work nobody has looked at, published on the site or not
- A time window, like “someone can be out there this afternoon”
- A warranty length, or telling a caller that a repair is covered
- “The estimate is free” when the trip charge is not
- Financing terms, or a senior or military discount that ended last year
Whether an agent should ever say a number at all is its own question, and it gets its own answer over on can an AI agent give a customer a price.
What to do next
None of this needs a purchase. It needs an afternoon and a written list.
- Write down what your chat is allowed to say about price, timing, warranty, and coverage. One page is enough. Anything not on that page becomes a handoff to a person.
- Ask your own chat those questions yourself, in the words a customer would use, before a customer does it for you. There is a fuller run through in how to test your website AI agent before you turn it on.
- Check the chat’s answers against your own pages. The tribunal in the Air Canada case refused to treat a webpage as inherently more trustworthy than a chat window, so a contradiction between the two helps nobody.
- Keep the transcripts. The screenshot that customer saved is quoted in the decision at paragraph 15.
- Decide in advance who calls it when the chat gets something wrong, and what the customer gets.
If you cannot yet write down what your agent may promise, do not turn one on. A chat that will answer anything is not an asset, and no monthly plan fixes that for you, whatever it costs per conversation. The setup described on the AI assistant page is built the other way around, so anything it should not decide gets handed back to a person.