Florida Man Pleads Guilty After OpenAI Reports Murder Plot Via ChatGPT

A 25-year-old former Goldman Sachs analyst has pleaded guilty to felony stalking and threat charges stemming from conversations with OpenAI's ChatGPT, where he detailed plans to harm his ex-girlfriend. OpenAI's internal safety team detected the concerning pattern and proactively notified the FBI.

The man received eight years of probation and two years of electronic monitoring as part of a plea agreement. He pleaded guilty to aggravated stalking, written threats to kill, and illegal use of a cell phone. The judge withheld formal adjudication, meaning the man will avoid a felony conviction if he successfully completes his probation.

OpenAI's policies include a review process for conversations that suggest users may be planning harm to others. In serious cases, the company contacts law enforcement. Federal agents later provided the Palm Beach County Sheriff's Office with approximately two months of chat logs, which reportedly showed not just venting but clear planning and rehearsal.

According to court records, the man repeatedly told ChatGPT he intended to kill his ex-girlfriend, describing specific scenarios and mentioning the purchase of firearms. His attorney later stated he did not possess any guns. Investigators also found that the man had used multiple app-generated phone numbers to send harassing and threatening messages to his ex-girlfriend.

This case is among a growing number where AI chatbot logs have become evidence in criminal investigations. It is, however, less common for companies like OpenAI to proactively report such findings to law enforcement, as opposed to having logs subpoenaed later in an investigation. This contrasts with a previous incident where OpenAI did not notify Canadian authorities about a user's violent intent, which later resulted in a mass shooting.