AI listened to the murder plan — and alerted the FBI

Darren Zhou, a 25-year-old from Florida who worked as an analyst at the investment bank Goldman Sachs, was arrested and prosecuted after sharing his detailed plans to rape and murder his former partner with ChatGPT. OpenAI, the company that develops the chatbot, identified the immediate danger and reported him to the FBI.

According to court records, Zhou turned to the artificial intelligence software last March to vent his frustration after breaking up with his partner of the last six months. He detailed to the chatbot her hobbies, the places she usually spends time, and his jealousy of the men around her. Soon, the conversations turned into concrete threats of rape, murder, and suicide.

"I am going to kill her by the end of the month," Zhou wrote to ChatGPT in conversations documented in the investigation file. "If I can't have her, then no one will."

Not an "emotional outburst" — premeditated planning

OpenAI's safety policy states that a small human team is authorized to review user interactions with ChatGPT in cases where the system detects an intent to harm others. In serious cases, the company transfers the information to law enforcement. While conversations with the chatbot have been used as evidence in courts before, cases where the company itself initiates contact with the police are quite rare.

In May, federal agents transferred Zhou's chat logs (without the bot's responses) to the Palm Beach County Sheriff's Office after the threats expanded to include the 21-year-old woman's family. An investigator who reviewed the correspondence determined that these were not "vague emotional outbursts," but a clear pattern of planning and rehearsals for committing a crime.

During a welfare check conducted by police, the young woman said she had broken up with Zhou due to his "unpredictable, jealous, and controlling behavior." Since the breakup, he had obsessively harassed her with calls, text messages, and on social media, using fictitious numbers. Many of the messages were sexual, insulting, and threatening in nature.

Avoided prison at the last moment

With the AI logs and threatening text messages in their possession, police arrested Zhou. He was fired from his job, spent two days in custody, and was released on a $100,000 bond.

In June, he was charged with a series of serious offenses — stalking with aggravating circumstances, written threats to kill, and illegal use of a cell phone — crimes that carry a maximum penalty of 25 years in prison. However, this week (August 13), Zhou signed a plea deal in which he admitted to all charges but avoided serving actual prison time and a serious criminal record.

Judge Scott Suskauer noted in the courtroom that he agreed to approve the deal, which includes eight years of probation and a series of strict restrictions, only because Zhou's former partner approved the arrangement. Zhou's attorney, Steven Bell, emphasized the young man's clean record and the fact that he did not possess weapons despite his threats:

"He is not a violent or bad person. He went through a very difficult mental health episode, and he is now doing extraordinarily well."