When a chatbot becomes a legal problem, the evidence question is whose record holds up

The chatbot made a promise. Now what?

A customer asks a chatbot about bereavement fares and gets an answer. Months later, they claim the airline owes them money based on what the chatbot said. The airline argues it can't be held responsible for information from its chatbot, as if the chatbot were a separate legal entity. The court in British Columbia disagrees: the airline is responsible for everything on its website, whether it comes from a static page or a chatbot ^[1].

That 2024 ruling in Moffatt v. Air Canada is one of several establishing that when a chatbot speaks on your website, it speaks for you. In New York, a car dealer's chatbot offered to sell a new vehicle for one dollar and threw in a free iPad. The dealership later tried to walk it back. A judge ruled in 2024 that the dealership had to honour the terms the chatbot had agreed to ^[2]. In Texas, a funeral home's chatbot gave advice on embalming that turned out to be wrong and potentially harmful. A court said the funeral home could be liable ^[2].

Think of it like this: if you put a sign in your shop window, you're responsible for what the sign says, even if you didn't write it yourself. A chatbot is the same.

First Amendment defences are failing

Some companies have tried arguing that forcing them to answer for every chatbot statement would violate their right to free speech under the First Amendment. Courts are not buying it. As one ruling put it, the First Amendment does not grant businesses the right to lie without consequence ^[2]. The principle is straightforward: customer service is commercial speech, and there are limits to how much you can mislead people in the course of selling something.

The evidence is often a screenshot

In the Air Canada case, the customer had a screenshot of the chatbot's answer, and an airline representative had already admitted the words were misleading. On the fare itself, the airline offered no evidence to the contrary, so the tribunal assumed such evidence wouldn't have helped it ^[1].

That is the pattern. The customer keeps a screenshot. The company has nothing to show what the chatbot actually said, or claims it can't retrieve the conversation. When the case reaches a courtroom or a tribunal, the side without its own records starts on the back foot.

In the EU, transparency rules apply from August 2026

Under Article 50 of the EU AI Act, which has applied since 2 August 2026, people interacting with an AI system must as a rule be informed that they are doing so, unless it is obvious from the context ^[3]. The rule covers customer service chatbots. It does not say how to inform people, but it must be clear and timely.

The transparency duty is separate from the question of liability. Telling someone they are talking to an AI does not, on its own, let you off the hook if the AI gives them bad information. What it does is set a baseline: people should know when they are not talking to a person.

What this means for you

If you run a chatbot that talks to customers, three questions are worth asking now:

1. Can you show what your chatbot said in a specific conversation six months ago, if someone disputes it?
2. Do customers know they are talking to an AI, clearly and early in the conversation?
3. If you had to defend what the chatbot promised in court, whose record would you rely on?

The law does not yet require you to seal or archive every chatbot conversation, but these cases show that not having your own evidence is a problem when things go wrong.

Sealing each answer gives you your own record

This is a case where a digital witness fits well. The thing to seal is each answer the chatbot sends, or each full conversation when it ends, including the notice that the customer was talking to an AI. Sealing takes a fingerprint of the answer, not the answer itself, so the content stays in your systems. Anyone you share the record with can later check that it is the one from that day, unchanged.

Who can check it? Customer service, the customer, an ombudsman or a court. What it shows is that the record of what the chatbot said, and when, has not changed since it was sealed. What it does not do is show that the answer was correct, or settle whether the company is bound by it. That is for a court to decide, but at least you have evidence.

You can seal answers through WitniumChain with the REST API, the SDK or the MCP server. The content never leaves your infrastructure. The first 2,000 witnesses a month are free. Read the documentation to see how it works.

How it would work

1. The chatbot sends an answer. Your system calculates a fingerprint and sends it to WitniumChain, getting a witness ID back.
2. The record (answer text, timestamp, conversation ID and witness ID) is stored in your own database.
3. Months later, if the conversation is questioned, you retrieve the record, calculate the fingerprint again and check it against the sealed one. If they match, nothing has changed.

What to watch

As these cases build precedent in the US, organisations elsewhere are paying attention. The EU's transparency rule is recent, and it will be a few years before we see how courts in Europe treat disputes over chatbot statements. The pattern from America suggests that having your own evidence, sealed at the time, will matter.

If you are about to write a tender or a policy on AI in customer service, one line is worth adding: the organisation must be able to retrieve and verify records of chatbot conversations.