"The original case involved a man suing an airline over an alleged personal injury. His legal team submitted a brief that cited several previous court cases in an attempt to prove, using precedent, why the case should move forward.

But the airline's lawyers later wrote to the judge to say they could not find several of the cases that were referenced in the brief.

"Six of the submitted cases appear to be bogus judicial decisions with bogus quotes and bogus internal citations," Judge Castel wrote in an order demanding the man's legal team explain itself."

BBC

Now, I have come across a case of a similar case at a large global insurer so this may be the tip of the iceberg. A long-time practitioner in AI, Christopher Surdak, brought the article to my attention when we were reviewing LLMs. In his own words: -

"Stunning. Nearly a billion people are using chatGTP without understanding what it actually is and does. They think it's a system that is aware of facts, rather than merely a system that applies statistical correlations between words as they appear in the English language. When an attorney asks it about case history, it has no idea what is meant by "case history," It only knows the statistical correlation between the words "case" and "history", along with the other words the attorney uses to perform a search for "case history." Hence, chatGPT will make up entirely fictional "case history" because none of what it is correlating has anything to do with actual cases or actual facts."

No wonder Chris is retained by organisations, governments, and enterprises to warn of the pros and cons of LLMs, chatGPT and, competing products like BARD and plug-ins. linkedin.com/in/csurdak

Best to know who is using it for what and when across your organisation. 

Further Reading

Insurer Zurich experiments with ChatGPT for claims and data mining

Navigating AI usage in the insurance sector

Part 1: Chatpocalypse Now

Part 2- Chat-tastic Results