November 4, 2026

"The Bot Said So": What Your AI Says, Your Company Said

Every company that puts a chatbot in front of customers is, in effect, hiring a new spokesperson. It works around the clock, answers instantly, and never needs a break. It also has no idea what the company's policies actually say unless someone makes sure it does.

When that spokesperson gets it wrong, who is responsible? The answer from tribunals and regulators is now consistent. Whatever a company's AI says to customers or investors, the company has said; regulators and tribunals are treating chatbot promises and "AI-washing" claims as the company's own representations.

There are two sides to this. The first is what the AI says. The second is what the company says about its AI.

What the AI says: Moffatt v. Air Canada

The leading case is small in dollars and large in principle. A customer asked Air Canada's website chatbot about bereavement fares. The bot misstated the airline's policy, and the customer relied on what it told him.

Air Canada argued, in substance, that the chatbot was a separate entity and the airline was not responsible for what it said. The British Columbia Civil Resolution Tribunal rejected that argument in 2024 and held the airline liable for negligent misrepresentation, awarding about C$812 [1]. The bot was not a separate entity. It was Air Canada.

The amount does not matter. The reasoning does. A business cannot deploy a tool to speak for it and then disown what the tool said. The research behind this article found no U.S. contract case directly following Moffatt yet, but the principle is the ordinary law of representations applied to a new speaker.

What the company says about its AI: AI-washing

The second line of cases runs the other way. Here the problem is not a chatbot's promise but the company's own claims about what its AI can do.

• Investment advisers. In March 2024, the SEC settled with Delphia (USA) and Global Predictions for false claims about using AI in investing, with penalties of $225,000 and $175,000 [2]. These were the first "AI-washing" cases.
• A public company. In January 2025, the SEC issued a cease-and-desist order against Presto Automation, which had overstated its AI drive-thru product and concealed how much of the work was done by people [3]. Statements about AI in public filings are in scope.
• A "robot lawyer." The FTC's final order against DoNotPay in February 2025 found the company could not substantiate claims that its service matched human lawyers. It paid $193,000 and had to notify subscribers [4].
• An accuracy claim. In August 2025, the FTC finalized an order against Workado, whose AI-content detector claimed 98% accuracy without support [5].
• Earnings claims to small businesses. Also in August 2025, the FTC sued Air AI, alleging deceptive claims about its AI "agents" and refunds. That case is pending, and the claims are allegations [6].
• Healthcare. In 2024, a health-AI company settled with the Texas Attorney General over alleged overstatement of the accuracy of its clinical summaries, agreeing to five years of disclosure obligations, with no penalty [7].

The pattern holds across federal and state regulators and across administrations. A claim about AI is a claim like any other. It needs support, and the company owns it.

Where the limits are

Not every AI statement becomes liability. In Walters v. OpenAI, a Georgia case, a radio host sued after ChatGPT falsely said he had been accused of embezzlement. In May 2025, the Gwinnett County Superior Court granted summary judgment to OpenAI: the output was not defamatory as a matter of law, and there was no fault and no damages [8]. Disclaimers, and what the user knew about the tool's limits, mattered.

That is a useful counterweight, but it is a defamation case about a general-purpose tool and a user who knew the tool's limits. It says little about a company's own customer-service bot answering a question about the company's own policy. And the questions are still being litigated: in Wolf River Electric v. Google, now pending in Minnesota state court, a business alleges that an AI search summary falsely said the state attorney general had sued it [9].

A general counsel's checklist

From the in-house side, the response is familiar. It is the same discipline companies apply to advertising, disclosures and investor communications, extended to a new channel.

• Treat the chatbot as a spokesperson. Limit it to approved policies and content, and test it against them before launch and after every change.
• Route the high-stakes questions to people. Refunds, pricing, eligibility, warranties and anything that sounds like a promise should go to a human or to a fixed, approved answer.
• Keep the logs. If a customer says the bot promised something, the company will want to know what it actually said.
• Run AI marketing claims through disclosure review. Accuracy rates, capabilities, and "AI-powered" labels need substantiation in the file, just as any performance claim does.
• Align the public filings. What the 10-K says about AI should match what the product actually does.
• Do not count on disclaimers alone. Walters shows they can matter. Moffatt shows they will not rescue a company from its own bot's answer on its own policy.

When the dispute comes

Most chatbot disputes will be small: a fare, a refund, a warranty. They are exactly the kind of dispute that should not take years. A clear dispute-resolution clause, with a mediation step and a fast arbitration track, lets a company resolve them quickly and privately and fix the bot before the next customer asks the same question.

The bigger lesson is simpler. Your AI is not a separate entity. When it speaks, you have spoken.

About the author: Marco G. DeSanto is the founder of DeSanto Arbitration. He is an attorney licensed in Michigan and is based in Atlanta, Georgia. As a neutral, Marco G. DeSanto does not represent or advise any party.

Sources

1. Moffatt v. Air Canada, 2024 BCCRT 149 (B.C. Civil Resolution Tribunal, Feb 14, 2024). https://www.canlii.org/en/bc/bccrt/doc/2024/2024bccrt149/2024bccrt149.pdf
2. SEC, Delphia (USA) Inc. and Global Predictions Inc., Advisers Act Rel. IA-6573, IA-6574 (Mar 18, 2024). https://www.sec.gov/newsroom/press-releases/2024-36
3. SEC, In re Presto Automation Inc., Securities Act Rel. 33-11352 (Jan 14, 2025). https://www.sec.gov/enforcement-litigation/administrative-proceedings/33-11352-s
4. FTC, DoNotPay final order, FTC Matter 2323042 (Feb 11, 2025). https://www.ftc.gov/news-events/news/press-releases/2025/02/ftc-finalizes-order-donotpay-prohibits-deceptive-ai-lawyer-claims-imposes-monetary-relief-requires
5. FTC, Workado LLC final order, FTC Matter 2323092 (Aug 28, 2025). https://www.ftc.gov/news-events/news/press-releases/2025/08/ftc-approves-final-order-against-workado-llc-which-misrepresented-accuracy-its-artificial
6. FTC v. Air AI Technologies (D. Ariz., filed Aug 25, 2025; pending). https://www.ftc.gov/news-events/news/press-releases/2025/08/ftc-sues-stop-air-ai-using-deceptive-claims-about-business-growth-earnings-potential-refund
7. Texas Attorney General, Pieces Technologies Assurance of Voluntary Compliance, DC-24-13476 (Sept 18, 2024). https://www.texasattorneygeneral.gov/news/releases/attorney-general-ken-paxton-reaches-settlement-first-its-kind-healthcare-generative-ai-investigation
8. Walters v. OpenAI, L.L.C. (Gwinnett Cty. Super. Ct., Ga., May 19, 2025), order granting summary judgment. https://lawrpg.org/2025/ai-and-the-law/cases/walters/walters-summary-judgment-order.pdf
9. Wolf River Electric v. Google (Ramsey Cty., Minn.; remanded from D. Minn. Jan 2026; pending), as reported by Reason/Volokh Conspiracy (Jan 12, 2026). https://reason.com/volokh/2026/01/12/google-missed-key-deadline-in-suit-alleging-googles-ai-libeled-business-court-holds/

This article is general information only and is not legal advice.

Back to Insights