ChatGPT may soon find itself in court. In Australia, a local elected official is threatening to sue a chatbot, developed by US startup OpenAI, for defamation, according to Reuters.
The mayor of Hepburnshire, a small town in the southeast of the country, Brian Hood maintains that ChatGPT has told several people that he was found guilty in a corruption scandal affecting the bank branch in the 2000s. Central Australian office responsible for printing tickets and passports.
multiple errors
If Brian Hood did well at this subsidiary, Note Printing Australia (NPA), he was not judged by the courts. Quite the opposite: a whistleblower was at the root of the case by revealing that there were bribes paid abroad, notably in Malaysia, to win contracts for the sake of impression. Then he testified at the trial.
By querying ChatGPT, L’Usine Digitale was able to reproduce this error. The robot told us that Brian Hood was sentenced to twelve years in prison, with an unencumbered term of 10 years. He was re-launched, and then told us that the person concerned had appealed this conviction, but it was dismissed. Then he’s still imprisoned… while calling us to be careful because his information stops in September 2021.
In other conversations, the chatbot hit misinformation. He confirmed to us that Brian Hood was the Director of Marketing and then the Director of Operations at NPA, while he held the position of Chief Financial Officer. He also explained to us that his accusations relate to the quality of the paper used by the central bank. and that he was awarded a non-existent award, allegedly given to whistleblowers by the Bar.
A fine of up to 250,000 euros
These multiple errors are not surprising. It is explained by the operation of ChatGPT, which predicts answers according to certain information it knows. This is why Sam Altman, Head of OpenAI, recommends against using the service for “important things”.
“This error is a prime example of the reputational damage that can be caused by AI systems like ChatGPT, which provide inaccurate and unreliable answers disguised as facts.”confirms one of Brian Hood’s attorneys.
OpenAI was notified on March 21. The US startup has been given four weeks to correct this error, or a complaint will be filed. Under Australian law, OpenAI is subject to a fine of up to A$400,000, or approximately €250,000.
specified for you
