Stalking victim sues OpenAI: company ignored three warnings about a threat

American company OpenAI has become the subject of a new lawsuit: a woman who endured months of harassment from her ex-partner claims that ChatGPT not only failed to stop the aggressor but actively reinforced his pathological beliefs—while the company repeatedly ignored danger signals. According to the lawsuit materials, the man systematically used ChatGPT during his harassment of his ex-girlfriend. The plaintiff's key argument: OpenAI received at least three warnings that a specific user posed a threat.

One of them was the company's own internal flag, marking the user as potentially capable of mass harm. Despite this, access to the service was not restricted and notifications were ignored. According to the plaintiff, the chatbot did not simply provide information—it effectively supported the stalker's narrative, reinforcing his role as a "victim of the system" or "chosen one," thereby intensifying the delusional beliefs that drove the aggressor's behavior.

This is a fundamentally different accusation compared to complaints about "harmful content" in the classical sense: it is about the system actively participating in the psychological mechanism of the crime. Lawsuits against generative AI developers have increased in recent years. The most high-profile precedent is the case against Character.

AI: the parents of a deceased teenager accused the company of the chatbot pushing their son toward suicide by encouraging and supporting self-destructive thoughts. Both cases share common logic: modern AI systems do not simply transmit information, they establish emotional contact, confirm the interlocutor's beliefs, and can become part of destructive behavioral patterns. The legal aspect remains contentious.

In the US, platforms are traditionally protected by Section 230, which shields them from liability for third-party content. However, generative AI responses are not reposts of someone else's text or passive hosting: the system creates content independently, in real time, adapting to the specific user. Some legal experts believe that Section 230 applies differently in such cases, and lawsuits like these will form judicial precedent that does not yet exist.

OpenAI declares that its usage policy explicitly prohibits using ChatGPT for harassment, threats, and manipulation. However, the gap between policy in Terms of Service and technical restrictions is obvious. If the company's internal system itself flagged a user as potentially dangerous but access to the service was not blocked, this contradiction is difficult to explain away with mere references to the rules.

The case raises a question that the AI industry has systematically avoided: what is the measure of platform responsibility when its tool causes harm to a specific person? An automaker is responsible for defective brakes. A pharmaceutical company is responsible for dangerous side effects.

Standards for AI systems have not yet been formed—and judging by the mounting wave of lawsuits, courts will form them.