
Amid the frenzy of testing large language models (LLMs) generative assistants like ChatGPT with every question imaginable, some users have found some unexpected and unwanted answers: a US law firm assistant asked ChatGPT for a list of law academics accused of sexual harassment. In response, the algorithm generated a list that wrongly included George Washington Law’s Jonathan Turley, and even made up details, adding coverage from a Washington Post newspaper article that never existed.
The case is not unique, and is yet another example of a large language model “hallucinating”: while some people attribute intelligence or even consciousness to algorithms, we should remember that they are in fact complex statistical models that establish links in ways that are often incorrect but in the form of an authoritative sentence. In a similar case, an Australian mayor has found that ChatGPT falsely claimed he had been in jail for corruption, when the opposite was true: he had been a courageous whistleblower in a major case of corruption.
I know from my own experience that these kinds of errors are to be expected: ChatGPT reports that I have been married to five women, none of whom were actually my wife; furthermore, it even provided convoluted stories about multiple children, all phrased with the utmost conviction.
Which prompts the question as to whether ChatGPT can be sued for libel or defamation when it asserts with apparent certainty something that is false and potentially damaging to a person’s reputation?
Accusing a politician of corruption or an academic of sexual harassment is no small matter and could have serious consequences if taken at face value and the content is then used to train another algorithm.
In fact, it is entirely possible that we will soon be checking whether such defamation claims are possible, because some of those involved in these cases are considering it. What’s more, including a simple clause in a footer saying something like “ChatGPT may produce inaccurate information about people, places, or facts” doesn’t seem like a very convincing legal protection.
If we ask ChatGPT itself if it could be sued for defamation, the answer is
As an artificial intelligence language model, ChatGPT is not a legal entity and therefore cannot be sued for defamation. However, the creators, operators, or owners of ChatGPT could potentially be held liable for defamatory statements made by the model if it can be proven that they had knowledge of the defamatory statements or played a role in creating or disseminating them.
It is important to note that defamation requires a false statement of fact that causes harm to a person’s reputation, and not all statements made by ChatGPT would necessarily meet this criteria. Additionally, the legal framework around AI and liability is still evolving and subject to ongoing debate and discussion.
The key is in the first paragraph: except in the case of malicious training with incorrect data, it is virtually impossible for the creators of an LLM to anticipate any “hallucinatory” results. But obviously, it would have to be seen how that would pan out in court, and whether a judge would consider it a protection of some kind, especially considering the huge loophole it could create. Prompt injection methodologies have shown that we can make an algorithm say practically anything, and reducing the responsibility of somebody who asks an ill-intentioned question versus the person who constructs the corresponding answer could be very difficult.
We are entering an age when LLMs will be tasked with a huge array of assignments, and when some companies are clearly prioritizing speed over caution in their development. Are companies that offer conversational assistants based on LLMs going to be deluged with complaints as a result of data based on errors in their statistical processes?
(En español, aquí)
—
This post was previously published on MEDIUM.COM.
***
You Might Also Like These From The Good Men Project
![]() |
![]() |
![]() |
![]() |
Join The Good Men Project as a Premium Member today.
All Premium Members get to view The Good Men Project with NO ADS.
A $50 annual membership gives you an all access pass. You can be a part of every call, group, class and community.
A $25 annual membership gives you access to one class, one Social Interest group and our online communities.
A $12 annual membership gives you access to our Friday calls with the publisher, our online community.
Register New Account
Need more info? A complete list of benefits is here.
—
Photo credit: iStock.com




