ChatGPT Account of Alleged Stalker to Remain Blocked, Judge Says

OpenAI must maintain its suspension on the account of a man who allegedly used ChatGPT to harass and humiliate his ex-girlfriend, as the chatbot fueled his delusions about her, a California state court ruled.

Judge Harold Kahn of California Superior Court, San Francisco at an emergency hearing Monday entered a temporary restraining order requiring the account be locked until May 6 when the court will determine whether to issue a longer-term injunction.

The woman, who is proceeding anonymously in the case, sued OpenAI last week, claiming she “is in immediate danger” from her ex-boyfriend who was recently released from a mental health facility and will certainly turn back to ChatGPT to plan threats and violence.

OpenAI said it suspended the account on April 8 when it was first informed of the situation. The woman sued a day later.

Kahn also said OpenAI must provide copies of the man’s chat logs to the plaintiff’s legal team by April 17.

The woman’s attorney Jay Edelson of Edelson PC applauded the ruling in a statement, saying: “We’re profoundly grateful the Court acted so swiftly to protect our client during what has been the most frightening days of her life.”

“It took a San Francisco County judge less than one business day to do what OpenAI refused to do for months,” the statement said.

OpenAI didn’t immediately return a request for comment.

The complaint alleged OpenAI’s ChatGPT-4o model, released in 2024, drove the man’s delusions and mental health crisis and allowed him to target his ex-girlfriend by generating false clinical reports that portrayed her as abusive and dangerous. He then spread those false reports to her family, friends, and colleagues. The chatbot also encouraged him to send her overt death threats.

The complaint alleged OpenAI’s automated safety system flagged his account in August 2025 for “Mass Casualty Weapons” and deactivated his account. But the next day, a human safety team member reviewed the account, which contained chat logs with specific names of people he was targeting, and determined the suspension was a mistake, the lawsuit said.

OpenAI also failed to take action when the plaintiff submitted a notice of abuse to the company, the complaint said.

The stalker was arrested in January this year and charged with four felony counts of communicating bomb threats and assault with a deadly weapon, but was found incompetent to stand trial. He was ultimately released from a mental health facility due to a procedural failure by the state, the lawsuit alleged.

Mayer Brown LLP represents OpenAI.

The case is Doe v. OpenAI Found., Cal. Super. Ct., No. CGC26635725, 4/13/26.