The Chatbot Is Not a Third Party: What Moffatt, OLG Hamm and California’s Section 1714.46 Decide

What to stop assuming

Stop assuming that a customer-facing chatbot's statements can be disowned as the output of a separate system. In the three instruments reviewed here, that argument has been rejected in two decided cases and removed by statute in one US state. What none of them yet does is establish a general rule, and this brief keeps that distinction visible.

The decision that matters is not that the operator lost, but what the operator was not allowed to argue.

The three instruments

Moffatt v. Air Canada, 2024 BCCRT 149 is a decision of the British Columbia Civil Resolution Tribunal, a small-claims body, issued on 14 February 2024⁵⁶. The decision text itself could not be opened for this brief, because the sites that host it refused automated access. What follows rests on two practitioner accounts, one of which quotes the decision at length.

According to that account, Air Canada argued that it could not be liable for information provided by "one of its agents, servants, or representatives – including a chatbot". The tribunal treated this as a suggestion that "the chatbot is a separate legal entity" and called it "a remarkable submission". It added that the chatbot "is still just a part of Air Canada's website" and that it "makes no difference whether the information comes from a static page or a chatbot". It found that Air Canada "did not take reasonable care to ensure its chatbot was accurate", and so found negligent misrepresentation. The American Bar Association's account reports damages of approximately CAD 650, plus pre-judgment interest and tribunal fees⁶.

OLG Hamm, 12 May 2026, case 4 UKl 3/25, Verbraucherzentrale NRW e.V. v Aesthetify GmbH is a ruling of the Higher Regional Court of Hamm in an action by a consumer association¹². The company's chatbot told users that two named doctors were specialists in plastic and aesthetic surgery or aesthetic medicine, which the claimant said they were not. The court granted an injunction against those statements.

As reported in a practitioner note based on the court's press release, the defendant was held responsible even if it had trained and programmed the chatbot only on correct data, and the chatbot was not a "third party" for the purposes of the law². The court allowed a further appeal (Revision) to the Federal Court of Justice, citing new legal questions about attributing a chatbot's false statements. The ruling is therefore not final. One law-firm account describes the doctors personally, as managing directors, as the responsible parties⁷; the case title and the court's own reporting name the company as defendant, and this brief follows the latter. This brief draws on the press release and the practitioner note, not the full judgment.

California Civil Code section 1714.46 provides that, in an action against a defendant who "developed, modified, or used" artificial intelligence alleged to have caused harm, "it shall not be a defense, and the defendant may not assert, that the artificial intelligence autonomously caused the harm to the plaintiff"³. Subsection (c) preserves "any other affirmative defense, including evidence relevant to causation or foreseeability" and evidence of other parties' comparative fault. The page reproducing the code shows an update marker of 1 January 2026; the operative date was not confirmed from the enrolled bill.

What each one settles

Where it stops

Moffatt binds nobody beyond its parties. A different Civil Resolution Tribunal decision from the same day, Visscher v. Air Canada, 2024 BCCRT 146, shows a tribunal member describing an earlier tribunal decision as "not binding on me"⁴. That is a statement about that earlier decision, but it indicates how little weight these small-claims rulings carry as precedent. The Hamm ruling is a single appellate-level judgment with a further appeal open, and it applies German unfair-competition law to statements about the operator's own people. Section 1714.46 governs California civil actions only.

None of the sources reviewed records a US court deciding a claim over a chatbot's statements, or any regulator action on the point. Whether the same reasoning reaches agentic systems, or chatbots built and run by a third-party vendor, is untested in these instruments.

What this requires in practice

None of this is a compliance rule, and this brief is not legal advice. Read together, the instruments describe an exposure rather than a duty: where a chatbot speaks for the business, the business is likely to be treated as having spoken.

Two practical points follow from the sources. In Moffatt, the tribunal expected the operator to take reasonable care that its chatbot's answers were accurate, so the evidence worth being able to produce is testing and review of what the chatbot says about people, qualifications, prices and terms. In the Hamm case, the chatbot had been deactivated but the defendant had given no cease-and-desist declaration, and the injunction still followed². Switching a system off is not the same as undertaking not to repeat its statements.

What to do with this

Apply one test before deployment and after every change: would the business stand behind this sentence if it were published on its own website under its name? If the answer depends on the sentence having been written by a model, the exposure described above is already in play.

Correct one assumption now. Neither training the system on accurate data nor describing it as a separate tool has, so far, worked as a defence in the two decided cases. Watch the Federal Court of Justice, which is the first place a binding answer on attribution is likely to appear, and read the full texts of Moffatt and the Hamm judgment before relying on either.