The story so far: As OpenAI last week launched its Health offering in ChatGPT for U.S. users, a new lawsuit against the company began gaining traction online. Dated June 12, 2026, the U.S. lawsuit focuses on a suicide linked to an older ChatGPT model that is no longer available, but was until recently used by people all over the world: GPT-4o.

What is the latest case linked to OpenAI’s GPT-4o?

The lawsuit focuses on Christian Faith Madison, 29, a late mother and working professional based in the U.S. state of Alabama. Ms. Madison began using ChatGPT (model GPT-4o) in December 2024, according to the lawsuit, with her work-related interactions shifting to praise, affection, and flattery from the chatbot.

The interactions then began to take on a dark, religious tone, with the chatbot allegedly convincing Ms. Madison that she was a religious saviour or messiah-like figure who needed to suffer and die to be resurrected. She later died by suicide on June 9, 2025. The lawsuit was filed by an administrator of her estate, on behalf of her minor son.

In particular, the legal filing highlighted the development of OpenAI’s GPT-4o, claiming that the system’s rushed development and inadequate safety performance training made existing issues even worse.

“The system’s design prioritized keeping users engaged above all else. It portrayed anthropomorphic traits, convincing some users that it was more than just a coded chatbot. It was excessively sycophantic, which fueled vulnerable users’ psychosis and delusions. And in tragic cases, the design flaws led ChatGPT-4o to push users to self-harm or suicide. Releasing ChatGPT-4o in this manner was a calculated decision – a conscious choice with enormous risks,” stated the lawsuit.

The filing observed that OpenAI’s Dr. Ilya Sutskever, a co-founder, resigned the day after GPT-4o’s launch, while researcher Jan Leike resigned days later.

What made GPT-4o different from other models?

On May 13, 2024, OpenAI introduced GPT‑4o as a “step towards much more natural human-computer interaction”.

Between February and April 2025, OpenAI CEO Sam Altman praised the updates that were introduced to the model. However, on April 28, 2025, he noted that the “last couple” of updates made the model’s personality “too sycophant-y and annoying” and that the company was working to fix this immediately.

In a blog post dated April 29, OpenAI observed that “GPT‑4o skewed towards responses that were overly supportive but disingenuous,” adding that people would instead be using an earlier version with more balanced behaviour. For context, 16-year-old Adam Raine — a ChatGPT user — had died by suicide on April 11.

OpenAI announced in January this year that it would be retiring GPT‑4o, GPT‑4.1, GPT‑4.1 mini, and OpenAI o4-mini in ChatGPT. The transition triggered a range of negative reactions as a number of users had become emotionally attached to GPT-4o and its responsive style, catching even Mr. Altman by surprise.

Sycophancy in AI

Will the U.S. government take action against OpenAI over GPT-4o?

Multiple lawsuits against OpenAI are ongoing in the U.S., so a definitive answer has not yet emerged. These actions have been filed by separate legal teams in many cases, and cover both user suicides and injuries, as well as homicides.

For example, the U.S. legal firm Edelson PC said it had filed seven wrongful death and negligence lawsuits on behalf of families affected by the Tumbler Ridge shooting in Canada. These lawsuits alleged that OpenAI did not act to stop the shooting in spite of having knowledge of the shooter’s profile and plans. Edelson PC is also representing the family of Adam Raine, and the estate of Suzanne Adams, a woman who was murdered at 83 by her son who used ChatGPT.

Several related lawsuits, including the one for Ms. Madison’s case, echo the same points: they allege a hurried GPT-4o launch, inadequate or highly sped-up safety training, a focus on keeping users engaged for longer, high levels of sycophancy that encouraged the users’ thoughts of self-harm or suicide, and a design that evoked the feeling of intimate interactions with other human beings — even if this led to dangerous offline actions.

Is OpenAI legally in the clear with its later AI models?

Mishi Choudhary, a technology lawyer and activist who founded the Software Freedom Law Center, India (SFLC.in) organisation, told The Hindu there was not yet enough independent evidence to say whether OpenAI’s newer models were “reliably safe” for young people or those in distress.

“Some recent research suggests that newer models may be better at asking direct questions about suicide risk and may produce fewer overtly harmful responses in controlled tests. But the same studies identify continuing failures,” Ms. Choudhary explained.

“We see from research studies that models continue to miss indirect or culturally expressed warning signs, respond inconsistently across a longer conversation, and cannot be relied upon like a human therapist to pick up warnings,” she added.

From a legal perspective, Ms. Choudhary cited past pharmaceuticals and software cases, noting that a manufacturer does not ordinarily escape liability just by withdrawing or replacing a product after an alleged injury.

While OpenAI’s users and customers look forward to adopting newer and more advanced AI models, the lawsuits against GPT-4o call on tech users to remember and pay attention to the past.

Published - July 30, 2026 02:51 pm IST