ChatGPT accused a man of a crime in a case the chatbot had fabricated, according to the court filings. (Bloomberg)

OpenAI Sued After ChatGPT Implicates Man in Alleged Offense

Febin MathewJune 8, 23 AI, Artificial Intelligence, ChatGPT, OpenAI

OpenAI has faced issues with its generative AI model, ChatGPT, as it produced fake legal citations and wrongly accused a man of a crime, leading to a libel lawsuit against the company. This comes after a lawyer was also put in trouble due to fabricated cases generated by the AI model. Despite these challenges, many countries have introduced their national strategy documents on AI, indicating the widespread adoption of artificial intelligence.

What happened

According to a report from Gizmodo, the first lawsuit has been filed against OpenAI since its popular generative AI chatbot ChatGPT accused a man of embezzlement. Earlier this year, a reporter asked ChatGPT to provide a summary of a lawsuit called The Second Change Foundation v. Robert Ferguson, which the chatbot immediately agreed to. In its response, the chatbot specified that the case involved Mark Walters, a radio host accused of embezzlement from a Second Amendment foundation. The problem was, however, that none of this information was true, and ChatGPT had given false information because Walters never worked for the foundation at any point.

Now, OpenAI, the company behind ChatGPT, is suing Walters for defamation. The lawsuit was filed in Gwinnett County Superior Court on June 5, according to the report. “Every statement of fact contained in the summary about Walters is false,” read the suit, and Walters’ attorney said the chatbot “posted defamatory material about Walters.”

Hallucinations

According to the researchers, Walters was on the receiving end of a so-called hallucination, where artificial intelligence models such as ChatGPT and Google Bard fabricate information and present it as facts.

Speaking about the issue of hallucinations at IIIT Delhi today, Sam Altman, founder and CEO of OpenAI, said, “It takes us about a year to improve the model. It’s a balance between creativity and accuracy, and we try to minimize the problem. (At the moment) I trust the responses from ChatGPT the least of anyone of the rest on earth.”