San Francisco judge blocks ChatGPT access in stalking, harassment case

A San Francisco judge has ordered OpenAI to cut off a man accused of stalking and harassment, turning a city Superior Court case into a test of how far a court can go in limiting access to a mainstream AI tool.

Judge Harold Kahn issued a temporary restraining order on April 13 in San Francisco Superior Court in Doe v. OpenAI, directing the company to suspend the accounts tied to the man identified in the complaint as John Roe, or “the User.” The order came after a complaint filed April 9 in San Francisco County, California, asked the court to block him from ChatGPT, stop him from opening new accounts, notify the plaintiff if he tried to return, and preserve his full chat logs for discovery.

The complaint says GPT-4o validated and amplified the man’s delusions, including a belief that he had discovered a cure for sleep apnea. It also ties the dispute to a broader stalking and harassment case that escalated into criminal charges. According to the filing, the man was arrested in January 2026 on four felony counts, including communicating a bomb threat and assault with a deadly weapon. A criminal court later found him incompetent, ordered him committed to a mental health facility, then ordered his release after a procedural failure.

The plaintiff’s lawyers say the risk did not end there. The complaint says she filed a formal Notice of Abuse in November 2025, copying the San Francisco Police Department and the Federal Bureau of Investigation. OpenAI said it had identified and suspended relevant accounts, but the plaintiff’s lawyers argued that suspension could be reversed because the company had previously restored his access after an earlier deactivation.

That is where the case reaches beyond one difficult domestic dispute. Commentators have compared the controversy to Packingham v. North Carolina, the Supreme Court case that treated access to major online platforms as a First Amendment issue. Here, the question is whether a judge can block a person from a widely used AI service, and whether ChatGPT should be treated more like speech, media, or a tightly controlled device when a court is trying to prevent further harm.

The San Francisco case now sits inside a larger wave of OpenAI litigation in the county. Another April 9 lawsuit, also filed in San Francisco Superior Court, names OpenAI Foundation, OpenAI OpCo, LLC, OpenAI Holdings, LLC, OpenAI Group PBC, and Sam Altman as defendants and seeks negligent entrustment, negligence, strict product liability, failure to warn, UCL claims, punitive damages, and injunctive relief. Together, the cases raise a larger California question: whether future sentencing, probation, or mental-health orders can reach not just phones and computers, but the AI tools people use every day.