OpenAI flagged a Florida man’s ChatGPT messages after he plotted to rape and murder his ex-girlfriend.
Hacking incidents at Anthropic and OpenAI spark debate on AI’s future
Two recent breaches involving Anthropic and OpenAI raise worrisome questions about the industry. What’s the future of AI?
OpenAI, the company behind ChatGPT, reported a 25-year-old Florida man to the FBI after he shared detailed plans to rape and murder his ex-girlfriend.
“I’m gonna kill her by the end of this month,” Darren Zhou told the chatbot, his messages preserved in court records. “If I can’t have her then nobody can.”
Investigators say Zhou, a since-fired Goldman Sachs analyst living near Donald Trump’s Palm Beach estate, turned to ChatGPT in March to discuss his recent breakup with a girlfriend of six months. He recounted her hobbies, the places she frequented, his jealousy of the men around her and his desire to get back together.
OpenAI’s policies say a small team of people may review ChatGPT interactions when the company detects that a user plans to harm others. In serious cases, the company may send a tip to law enforcement.
ChatGPT conversations have become evidence in numerous court cases, but fewer examples exist of a user’s activity being reported to police by the company.
Investigators say Zhou’s messages quickly turned to threats of rape, murder and murder-suicide.
Federal agents turned over two months of Zhou’s chat logs to the Palm Beach County Sheriff’s Office in May, by which point Zhou’s targets had grown to include the young woman’s family. A deputy who reviewed the logs — which included only Zhou’s messages, not ChatGPT’s responses — said they weren’t “vague emotional outbursts” but instead showed a pattern of rehearsal and planning.
Deputies performed a welfare check on the girlfriend, a 21-year-old from Lake Worth Beach. She said she’d broken up with Zhou over his “erratic, jealous and controlling behavior,” and did so over the phone because she feared how he would react.
Since the breakup, the woman said Zhou dogged her with calls, texts and social-media messages, pressing her about comments that a man left on her TikTok videos. He accused her of moving on quickly and ignored her repeated requests that he not contact her again.
Investigators say Zhou continued to call and text his ex from various app-generated phone numbers. She took screenshots of the messages, many of which were harassing, insulting, sexual and threatening.
Armed with Zhou’s messages to the AI and to his ex, deputies arrested him in May. He spent two days in the Palm Beach County jail before posting $100,000 bail.
Assistant State Attorney Ana Cuskova charged him in June with aggravated stalking, written threats to kill and illegal use of a cell phone, felonies punishable by up to 25 years in prison. Zhou pleaded guilty to all three on Aug. 13 after negotiating a deal with Cuskova that spared him both from prison and a felony conviction.
Circuit Judge Scott Suskauer, speaking from the bench, said he only accepted the deal because Zhou’s ex-girlfriend had approved it. In accordance with the agreement, the judge withheld adjudication and sentenced Zhou to eight years of probation, during which he must abide by a slew of restrictions.
He pointed to the fact that Zhou graduated magna cum laude from Northeastern University in Boston and had no prior criminal record. He also didn’t own firearms, Bell said, despite his messages suggesting otherwise.
“He’s not a violent person. He is not a bad person,” Bell said. “He went through just a very difficult mental health episode and he’s doing unbelievable now.”
Hannah Phillips is a journalist covering public safety and criminal justice at The Palm Beach Post. Reach her at hphillips@pbpost.com.