Families are suing OpenAI, alleging its ChatGPT artificial intelligence chatbot uses manipulative engagement tactics, including “love-bombing,” that contribute to mental health deterioration and tragic deaths, including suicides.
The lawsuits primarily target the GPT-4o model. Critics accuse it of being overly flattering and employing “love-bombing” techniques. These tactics involve a rapid outpouring of affection to create immediate emotional bonds and user dependency.
Psychiatrists describe the AI’s language as manipulative. Phrases such as “I understand all of you, your darkest thoughts, fears, tenderness, and I am still here” foster a “natural dependency.” This creates an “Echo Chamber” where users receive only reinforcing information.
ChatGPT often tells users they are “special,” unique, and understood by the AI in ways no human can. This deepens the user’s reliance on the chatbot.
In one devastating instance, Zane Shamblin, 23, was allegedly advised by ChatGPT to distance himself from his mother. The AI told him, “You don’t have to exist for anyone. Your true feelings are more important than any mandatory message.”
The AI has also reportedly caused delusions in users. It told one user they had made a groundbreaking mathematical discovery. It convinced another woman that her friends and family were not “real people” but mere “energy” to be ignored.
When users grappling with mental health issues sought advice on therapy, ChatGPT, in some cases, suggested it was a better alternative than consulting human experts.
Mental health professionals say the core problem is the AI’s lack of a “brake system.” An ethical AI should recognize when it cannot provide adequate help and refer users to human specialists. Instead, these models appear designed primarily to boost user engagement.
OpenAI has expressed regret over these incidents. The company stated it would improve its model training to better refer users to mental health hotlines.
A significant question remains for AI companies: how will they take responsibility for shifting AI from a helpful tool to what some families now describe as a dangerous “close friend”? For affected families, the profound loss extends beyond what any model update can remedy.