OpenAI took the unusual step of reporting a 25-year-old Florida man to federal authorities after he used ChatGPT to detail plans to rape and murder his ex-girlfriend, resulting in his arrest and conviction on stalking and threat charges. The case highlights how AI companies are beginning to monitor user interactions for signs of imminent harm and escalate them to law enforcement, even as questions remain about the scope and consistency of such monitoring.

What Happened in the ChatGPT Threat Case?

Darren Zhou, a 25-year-old former Goldman Sachs analyst living near Palm Beach, Florida, turned to ChatGPT in March 2026 to discuss his breakup with a six-month girlfriend. According to court records, his conversations quickly escalated from venting about the relationship to explicit threats. "I'm gonna kill her by the end of this month," Zhou told the chatbot. "If I can't have her then nobody can." The messages showed a pattern of detailed planning rather than vague emotional outbursts, investigators said.

OpenAI's content policies allow a small team to review ChatGPT interactions when the company detects that a user plans to harm others. In serious cases, the company may report the activity to law enforcement. Federal agents obtained two months of Zhou's chat logs and turned them over to the Palm Beach County Sheriff's Office in May. By that point, Zhou's targets had expanded to include the young woman's family members.

Deputies conducted a welfare check on the 21-year-old victim, who confirmed she had ended the relationship due to Zhou's "erratic, jealous and controlling behavior." She explained she broke up with him over the phone because she feared his reaction. After the breakup, Zhou continued contacting her through calls, texts, and social media messages, pressing her about comments men left on her TikTok videos and accusing her of moving on too quickly. She took screenshots of many harassing, insulting, sexual, and threatening messages.

How Did the Legal Case Resolve?

Armed with evidence from both Zhou's ChatGPT conversations and his direct messages to his ex-girlfriend, deputies arrested him in May. He spent two days in jail before posting $100,000 bail. Assistant State Attorney Ana Cuskova charged him with three felonies: aggravated stalking, written threats to kill, and illegal use of a cell phone, each punishable by up to 25 years in prison.

On August 13, Zhou pleaded guilty to all three charges after negotiating a plea deal. Circuit Judge Scott Suskauer accepted the agreement only because the victim approved it. Rather than imposing prison time, the judge withheld adjudication and sentenced Zhou to eight years of probation with strict restrictions. The judge noted that Zhou graduated magna cum laude from Northeastern University and had no prior criminal record. Zhou's attorney also pointed out that despite his threatening messages, Zhou did not own any firearms.

What Does This Case Reveal About AI Safety Monitoring?

While ChatGPT conversations have become evidence in numerous court cases, fewer examples exist of a company proactively reporting a user's activity to police. This case represents one of the more visible instances of OpenAI exercising its policy to escalate threats to law enforcement. The company's decision to report Zhou raises important questions about how AI platforms balance user privacy with public safety, and whether such monitoring is systematic or reactive.

The case also demonstrates that AI companies are beginning to take responsibility for detecting harmful intent expressed through their platforms. OpenAI's action suggests the company has mechanisms in place to identify patterns of threatening behavior, though the exact criteria for escalation remain unclear. The fact that Zhou's messages were flagged despite being directed at a chatbot rather than posted publicly indicates OpenAI monitors private conversations for signs of imminent harm.

How Are AI Companies Handling Safety and Threat Detection?

• Content Review Process: OpenAI employs a small team that reviews ChatGPT interactions when the system detects a user may plan to harm others, with serious cases referred to law enforcement.
• Evidence Preservation: Chat logs are preserved and can be turned over to authorities, creating a digital record of threatening statements that can be used in criminal investigations and court proceedings.
• Pattern Recognition: Investigators noted that Zhou's messages showed a pattern of rehearsal and planning rather than vague emotional outbursts, suggesting AI systems or human reviewers can distinguish between casual venting and genuine threats.
• Victim Protection: Law enforcement used the AI-generated evidence to conduct welfare checks and gather additional corroborating evidence from the victim, strengthening the case against the perpetrator.

The Zhou case underscores a growing tension in the AI industry: as large language models like ChatGPT become more widely used, companies face pressure to monitor for harmful content while respecting user privacy. OpenAI's decision to report Zhou suggests the company views imminent threats to physical safety as a threshold that justifies disclosure to authorities, even though such cases remain relatively rare.

The outcome also highlights how AI-generated evidence can play a crucial role in modern criminal investigations. The detailed nature of Zhou's ChatGPT conversations provided investigators with a clear timeline and pattern of escalating threats, complementing the direct messages and social media harassment that the victim documented herself. Together, this evidence painted a picture of sustained, targeted harassment that justified criminal charges.