Darren Zhou, a 25‑year‑old former Goldman Sachs analyst, pleaded guilty on August 13, 2026, to three criminal charges in Palm Beach County, Florida, after a two‑month streak of threatening his ex‑girlfriend on ChatGPT. The FBI, acting on a notification from OpenAI, seized the chat logs and used them to build the case that led to Zhou’s arrest in May.
Zhou’s conversations, which began in March 2026 following their breakup, evolved from jealousy to explicit plans to kill, sexually assault, and hold his ex hostage. The logs reference an AR‑15‑style rifle, a Glock handgun and a 12‑gauge shotgun, details that were extracted from court filings.
The charges brought against him were aggravated stalking, written threats to kill and illegal use of a cellphone. Judge Scott Suskauer, who heard the plea, dismissed the adjudication and imposed an eight‑year probation order that includes electronic monitoring, a batterer‑intervention program, a mental‑health evaluation and a strict no‑contact rule with the former girlfriend.
The ex‑girlfriend, 21, told investigators she ended the relationship because of Zhou’s controlling behaviour and had been threatened after the breakup. She confirmed receiving repeated calls, texts and social‑media messages even after she blocked him, and provided screenshots that corroborated the chat history.
OpenAI’s decision to forward the logs to the FBI underscores a growing trend of tech firms cooperating with law‑enforcement when user content poses imminent threats. Similar cooperation was seen in a Bengaluru triple‑murder investigation where the accused used Google’s Gemini chatbot to plan the killings.
Zhou will now be monitored under the probation terms, with the FBI retaining the authority to step in should the conditions be violated. The case marks the first instance of an AI‑derived threat being prosecuted in a U.S. court.