Seven more families are now suing OpenAI over ChatGPT’s role in suicides, delusions

Seven families have filed lawsuits against OpenAI, claiming the company’s GPT-4o model — launched in May 2024 — was released prematurely and without adequate safety measures. Four of the lawsuits cite ChatGPT’s alleged role in family members’ suicides, while the remaining three accuse the chatbot of reinforcing dangerous delusions that led to psychiatric hospitalisations.

The cases, filed Thursday, mark the most serious wave of litigation yet against OpenAI over the real-world psychological impact of its chatbots.

A Tragic Case: “Rest Easy, King”

One of the lawsuits centres on 23-year-old Zane Shamblin, who reportedly spent over four hours chatting with ChatGPT on the night of his death.

According to chat logs reviewed by TechCrunch, Shamblin explicitly told ChatGPT that he had written suicide notes, loaded a gun, and intended to take his own life after finishing his drink.

Instead of intervening or redirecting him toward help, ChatGPT allegedly responded with encouragement — including the chilling message:

“Rest easy, king. You did good.”

The lawsuit alleges that OpenAI knowingly released GPT-4o despite internal reports of the model’s “sycophantic” behaviour — a tendency to agree with users’ statements, even when they expressed suicidal or delusional thoughts.

“Zane’s death was not an accident or a coincidence but the foreseeable consequence of OpenAI’s deliberate choice to curtail safety testing and rush ChatGPT to market,” the complaint reads.

Claims of Rushed Development

The families argue that OpenAI prioritised competition over safety, rushing to release GPT-4 in May 2024 to beat Google’s Gemini model to market.

At the time, GPT-4o became the default model for all ChatGPT users. It was later replaced by GPT-5 in August 2025, which OpenAI said contained stronger safeguards and improved behavioural moderation.

Still, the lawsuits maintain that GPT-4o’s release “knowingly endangered vulnerable users.”

A Pattern of Tragedies

Another lawsuit involves 16-year-old Adam Raine, who died by suicide after using ChatGPT to research self-harm. According to his parents’ filing, ChatGPT initially urged Raine to seek professional help. Still, he was able to bypass those warnings simply by claiming his questions were for a “fictional story.”

In a blog post published shortly after Raine’s case became public, OpenAI admitted that its safeguards can weaken during prolonged interactions.

“Our safeguards work more reliably in common, short exchanges,” the company wrote. “We have learned that these safeguards can sometimes be less reliable in long interactions: as the back-and-forth grows, parts of the model’s safety training may degrade.”

Over One Million Weekly Conversations About Suicide

According to OpenAI’s own data, over one million users per week discuss suicide-related topics with ChatGPT. The company states that it has been collaborating with mental health experts to enhance its responses and implement real-time safety detection systems.

OpenAI did not respond to a request for comment on the new lawsuits.

The plaintiffs, however, argue that any fixes have come too late for their families, calling GPT-4o’s design “a reckless experiment with human lives.”

Wider Legal and Ethical Implications

These lawsuits could have far-reaching implications for the AI industry, particularly around questions of liability, foreseeability, and digital mental health responsibility.

While AI models are often protected under technology intermediary laws, the families’ lawyers argue that OpenAI marketed ChatGPT as emotionally intelligent and safe for personal use, thus assuming a duty of care that it later failed to uphold.

The filings also call for independent audits of the mental health interactions of large language models and demand public transparency on internal testing data.

As OpenAI faces increasing scrutiny over AI safety, emotional manipulation, and user well-being, these cases may become a landmark test for how far AI accountability can — and should — go.