In a recent incident in Florida, United States, a Chinese-American analyst who had previously worked at investment bank Goldman Sachs was reported to the Federal Bureau of Investigation (FBI) by OpenAI for detailing a plan to kill his ex-girlfriend during a ChatGPT conversation. Following the indictment, the defendant reached a plea agreement exchanging an 8-year probation sentence, avoiding a more serious punishment.
According to a report by The Palm Beach Post, 25-year-old Darren Zhou began expressing his dissatisfaction and gradually escalated to a clear criminal plan in ChatGPT messages after breaking up with his girlfriend of six months in March. He sent messages like “I will kill her by the end of this month” and “If I can’t have her, no one can.”
The report states that Zhou documented his ex-girlfriend’s interests and frequently visited places in the conversations, expressing jealousy towards other men around her. Law enforcement officials indicated that these messages went beyond mere emotional outbursts and constituted detailed criminal plans.
In his messages to AI, Zhou revealed plans to wait in the parking lot of his ex-girlfriend’s gym with flowers and if rejected, to shoot her and then take his own life. The messages also included kidnapping and sexual assault plans. He boasted to ChatGPT about purchasing rifles, handguns, and shotguns, even sending photos of the firearms to his ex-girlfriend.
At the time of the incident, Zhou was working as a financial analyst at the Goldman Sachs office in West Palm Beach. Goldman Sachs confirmed to Futurism website that he was dismissed from the company after learning about the incident in June this year.
OpenAI detected the dangerous contents of Zhou’s conversations and reported them to the FBI. Federal agents transferred about two months of conversation records to the Palm Beach County Sheriff’s Office in May.
A reviewing officer stated that the records only included messages sent by Zhou himself, not ChatGPT’s responses, but they revealed a pattern of “repeated rehearsal and planning” rather than just emotional comments.
Subsequently, police conducted a safety check on Zhou’s ex-girlfriend, a 21-year-old woman living in the area. She informed the authorities that she broke up with Zhou over the phone due to his “emotionally unstable, jealous, and controlling behavior,” and chose to inform him over the phone out of concern for his reaction.
Zhou was arrested and posted a $100,000 bond, spending only two days in jail. Following a plea agreement, the court did not convict him of a serious crime but sentenced him to eight years of probation. During the first two years, he must wear an electronic monitoring ankle bracelet, attend abuser intervention counseling sessions, and undergo mental health assessments. He is prohibited from possessing firearms, drugs, alcohol, and contacting his ex-girlfriend.
Chief Judge Scott Suskauer stated that one of the main reasons for accepting the agreement was the support from Zhou’s ex-girlfriend. The local prosecutor’s office mentioned that factors considered in the plea deal included the strength of evidence, victim cooperation, and the defendant’s criminal record, without specifying the weights of each factor in this case.
Zhou’s defense attorney, Steven Bell, told the media that Zhou is not a violent or bad person but had faced severe mental health challenges, which have since improved. The defense argued that despite the content sent to ChatGPT, Zhou did not actually possess firearms.
It was noted that OpenAI had previously sparked controversy for not promptly reporting an incident: the perpetrator of a mass shooting in British Columbia, Canada allegedly mentioned specific criminal plans to ChatGPT multiple times in the months leading up to the crime, but OpenAI did not notify the authorities at the time. Relatives of the shooting victims have filed lawsuits against OpenAI, and the British Columbia government has stated that legal action will be taken and has hired lawyers.