I recently had a lively online discussion about chatbots and liability, and it quickly emerged that one source of disagreement was that many people in the US tend to think that US precedents are all that matter. And this is obviously not the case. For example, the nearest thing to black letter law is the US about LLMs and copyright is Bartz v. Anthropic (2025), which rejected copyright infringement on the basis of “fair use”. (Yes, this is over-simplified.) But UK and EU law does not recognize “fair use” in the US sense.
So I decided to explore another kind of liability: defamation. I posed the following scenario to Copilot: (Cue ironic comments.)
- A user asks a chatbot about an author’s position on a topic.
- The chatbot hallucinates an inaccurate (and arguably defamatory) quote and generates a fake citation to a (real) book by the author.
- The chatbot provides the quote and citation to its user.
- The user then publishes this fake quote and citation in good faith.
- Can the author sue for defamation? If so, who is liable?
Copilot’s response was that the user might be held liable, but so far chatbot operators have been able to rely on CDA Section 230.
Could the AI company ever be liable?
Only in narrow scenarios:
- If the model was trained on defamatory content about the same person and reproduced it.
- If the company knowingly allowed the model to generate harmful falsehoods about identifiable individuals.
- If future legislation creates AI-specific liability.
Right now, courts have not imposed liability for hallucinated defamation.
But obviously this only applies in the US, so I asked Copilot to compare the US, UK and EU. The resulting analysis is lengthy (I invite you to try the same kind of prompt sequence), but the bottom line is pretty clear.
Your hypothetical — a hallucinated defamatory quote — is legally dangerous in the UK and EU in a way it is not in the U.S.
- In the U.S., the user is almost always the only viable defendant.
- In the UK, the AI company could realistically be sued as a publisher.
- In the EU, the DSA creates a path to liability because the AI company is the “content provider.”
This is why AI companies are far more cautious in Europe and the UK than in the U.S.
More anon.