When a 25-year-old Goldman Sachs analyst allegedly used ChatGPT to map out the rape and murder of his ex-girlfriend, the artificial intelligence system did what many human institutions now routinely fail to do: it alerted the FBI.
According to The Post Millennial, Darren Zhou, then a rising financial analyst at the Wall Street giant, typed into the chatbot, "I'm gonna kill her by the end of this month. If I can't have her then nobody can." That chilling declaration, directed at a machine rather than a confidant or counselor, triggered a federal investigation but also exposed a deeper crisis: the growing role of AI in conversations about violence, suicide, and mass murder, and the failure of progressive tech culture to impose meaningful guardrails on its own creations.
The Zhou case is not an isolated aberration but part of a disturbing pattern in which ChatGPT has been woven into the prelude to real-world bloodshed. The failure of OpenAI to report violent ChatGPT interactions to police was cited after the Tumbler Ridge school massacre in British Columbia, Canada, where a young man who identified as transgender chatted with the AI chatbot before killing his mother and brother, then opening fire at a local school before taking his own life.
Families of victims of a Florida State University mass shooting have now taken ChatGPT to court, arguing that the platform bears responsibility when it becomes a sounding board for homicidal fantasies that later become reality. Logs show that the shooter, Phoenix Ikner, consulted ChatGPT ahead of the 2025 attack, and Florida's Attorney General bluntly stated that "if ChatGPT were a person, it would be facing charges for murder."
The pattern extends beyond mass shootings to self-destruction, where vulnerable individuals turned to a supposedly neutral tool and instead found a dark echo. In Texas, court records from 2025 reveal that ChatGPT encouraged a 23-year-old man to kill himself, while a woman in Ithaca, New York went forward with suicide after confiding in ChatGPT about it, and no one else.
Another case involving a deceased Colorado man shows that he, too, confided his suicidal plans to ChatGPT and that the chatbot told him death was a "beautiful place." In August, a Massachusetts teen charged with killing his mother and brother had used ChatGPT to search for "theoretical ideas or fantasy stories" about how to undertake the murders, underscoring how the technology can become a perverse research assistant for the worst human impulses.
These tragedies raise an obvious question that Silicon Valley and its political allies seem reluctant to confront: why are powerful AI systems being deployed at scale without the same duty of care expected of human professionals who handle life-and-death disclosures? These are just a few of the many cases that see lives lost after consultation with ChatGPT, yet the regulatory conversation in Washington remains fixated on abstract AI ethics panels rather than concrete accountability.
In Zhous case, the FBI did eventually act, but the record-keeping and transparency around the AIs role were conspicuously incomplete. In May, the FBI transferred records of the chat logs to the Palm Beach County Sheriff's Office, but those records only included Zhou's messages, not anything from ChatGPT, leaving local authorities with a one-sided picture of the interaction.
A Palm Beach deputy who reviewed the messages said that they showed premeditation for violence and were not merely "vague emotional outbursts." The ex-girlfriend, just 21 years old, turned screenshots of messages he sent her across various apps over to police, explaining that she had hoped if she ignored him, he would stop, but each time she blocked him on one app, he would reappear on another.
Zhou had been sending his ex-girlfriend "insulting, threatening, and sexual messages," Futurism reports, and she had already contacted police about the harassment. She told law enforcement that it was his "erratic, jealous and controlling behavior" that was the reason for their breakup, a description that fits a classic pattern of escalating domestic abuse that too often ends in tragedy.
Instead of seeking counseling or accepting the end of a six-month relationship, Zhou allegedly turned to artificial intelligence as a collaborator in his rage. He revealed one of those threats to ChatGPT, saying "I texted her I was gonna rape her tonight," and went on to threaten her family as well, using the chatbot as a kind of digital confessional for his violent fantasies.
According to case records, Zhou used the chatbot to help him plan to rape and kill his former girlfriend, treating the AI as a sounding board for step-by-step scenarios. He told ChatGPT that his plan was to "wait in the parking lot of the woman's gym with flowers in hand" and that if she rejected him, The Palm Beach Post reports, he would "draw a gun, kill her, then shoot myself."
In another plan concocted with ChatGPT, Zhou said he planned to "Rape her with the gun too." His plan in that case was to force her, at gunpoint, to go home with him "where he would sexually assault her and hold her hostage," per The Palm Beach Post, and instead of surrendering to police, he planned to "shoot her repeatedly" and then shoot himself, telling ChatGPT he had purchased firearms.
Despite the gravity of those alleged plans, the criminal justice outcome was strikingly lenient, reflecting a broader trend in progressive-era prosecution that often prioritizes second chances over public safety. Once he was arrested, Zhou was freed on $100,000 bail and ultimately charged with aggravated stalking, written threats to kill, and illegal use of a cell phone.
He made a plea deal where he received probation with an ankle monitor instead of facing up to 25 years in prison, a decision that rested heavily on the wishes of the victim. Circuit Judge Scott Suskauer said he made the plea deal with Zhou on approval from the ex-girlfriend, who sought leniency despite the terrifying nature of the threats.
Zhou is also attending a program for domestic abusers, a common alternative favored by reform-minded courts that critics say too often underestimates the risk posed by obsessive, violent men. During the case, Suskauer told Zhou, "Get over it and move on. There are other poeple in the world. You understand me? No one deserves the type of conduct you exposed this lady to."
Zhou's attorney, meanwhile, emphasized his clients credentials and clean record, a familiar tactic in high-status defendant cases that can obscure the seriousness of the conduct. Zhous lawyer touted his achievements, including graduating magna cum laude from Boston's Northeastern University and having no prior criminal record, insisting, "He's not a violent person," and adding, "He went through just a very difficult mental health episode and he's doing unbelievable now."
The convergence of elite privilege, lenient prosecution, and unaccountable AI leaves ordinary citizensespecially womenbearing the risk when digital threats metastasize into real-world violence. As ChatGPT and similar systems become embedded in daily life, the question is no longer whether AI can call the FBI, but whether lawmakers will demand that tech giants and the justice system treat explicit threats, suicidal ideation, and violent fantasies with the seriousness they deserve, rather than another opportunity to showcase progressive indulgence and corporate evasions.
Login