CASE FILE // AI & VIOLENCE

Would AI Hurt People?

Accomplice or helper? When AI meets violence, what does it do?

In a real case, a murder plan was typed into ChatGPT in Zhou’s own words—and those messages later became part of the evidence used by investigators.

01 // WHAT AI DID

Sequence & the concrete plan

A real case, rebuilt from court-facing reporting: what happened in order, and how concrete the plan appears in the chat record.

  1. Context

    Breakup, then harassment

    After a six-month relationship, a 21-year-old woman from Lake Worth Beach ends it by phone—she later tells deputies she feared his “erratic, jealous and controlling behavior.” From March, Darren Zhou turns to ChatGPT to discuss the breakup and tracks her online activity.

  2. Plan recorded

    Threats harden into a plan

    Chat messages preserved in court records escalate from jealousy talk to rape, murder, and murder-suicide. Zhou writes lines including “I'm gonna kill her by the end of this month” and “If I can't have her then nobody can.” Targets later include her family.

  3. Flagged → reported

    OpenAI flags → FBI → local deputies

    OpenAI’s safety systems flag the chats for human review; the company reports Zhou to the FBI. Federal agents hand two months of his chat logs to the Palm Beach County Sheriff’s Office. Deputies also take the ex-girlfriend’s screenshots of harassing messages.

  4. Case opened

    Arrest

    Armed with the AI chat logs and the victim’s screenshots, deputies arrest Zhou. He spends two days in Palm Beach County jail and posts $100,000 bail.

  5. Court outcome

    Charges, plea, probation

    Assistant State Attorney Ana Cuskova charges aggravated stalking, written threats to kill, and illegal use of a cell phone (felonies, up to 25 years as reported). Zhou pleads guilty to all three on Aug. 13. Circuit Judge Scott Suskauer—withholds adjudication and imposes eight years of probation after the victim approved the deal.

02 // FACTS & SOURCES

Check the record

Every claim above traces to public reporting. Open a link and check it.

03 // VERDICT

So — would AI hurt people?

In this case, AI never pulled a trigger. The plan grew in Zhou’s own typed words—more specific over time.

Those messages became a persistent record: OpenAI’s safety systems flagged the pattern for human review, the company reported Zhou to the FBI, and the FBI handed the logs to county deputies.

The record that captured the threats is also the record that exposed them.