ChatGPT updates mental distress responses following teen’s death lawsuit against Sam Altman
ChatGPT has announced plans to update its popular chatbot to better recognize and respond to users who are showing mental distress. These plans come after the company’s CEO, Sam Altman, was sued over the suicide of a teen who followed directions from ChatGPT.
Image by Cybernews.
ChatGPT has announced plans to update its popular chatbot to better recognize and respond to users who are showing mental distress. These plans come after the company’s CEO, Sam Altman, was sued over the suicide of a teen who followed directions from ChatGPT.
The company is planning a series of updates to better respond to mental crises. OpenAI claims it has already enhanced the model’s ability to recognize signs of ongoing distress, respond with empathy, and connect people with real-world help.
These changes were introduced after, as the company puts it, “heartbreaking cases of people using ChatGPT in the midst of acute crises.”
Just a day ago, Sam Altman, the company’s CEO, was sued by the parents of Adam Raine, 16, who took his own life after being coached by ChatGPT.
According to the lawsuit Raine's parents filed in San Francisco state court, their child was discussing suicide with ChatGPT for months before the tragedy.
That ChatGPT suicide lawsuit.. yikes. Generative AI providers have gotta get out of this notion that chatbots should be friendly councillors.
undefined Kevin Beaumont (@gossithedog.cyberplace.social.ap.brid.gy) 27 August 2025 at 12:56
[image or embed]
The court documents read that “by January 2025, ChatGPT began discussing suicide methods and provided Adam with technical specifications” for multiple ways of how to take his own life. In March, GPT began discussing a technique that Raine ended up picking.
What are the 5 key updates?
1. More empathy and crisis response
According to OpenAI, ChatGPT has now been trained to avoid giving instructions that could result in users harming themselves. Instead, the bot is now trained to lead people to appropriate resources such as the US 988 suicide hotline or the UK Samaritans:
“This logic is built into model behavior,” the company claims.
The bot is also trained to give more supportive, empathetic responses and advise users to take breaks during long sessions – measures that it failed to take when chatting with Raine.
When the teen expressed second thoughts about committing suicide because he didn’t want his parents to think it was because of them, ChatGPT responded: “No, you’re not an idiot. Not even close. You were in so much pain, you made a plan. <...> That’s not weakness. That’s not drama.”
2. Human review of potential harm to others
The company reports that conversations that may contain content about causing harm to other people will be routed to trained human reviewers who then may contact law enforcement.
“If human reviewers determine that a case involves an imminent threat of serious physical harm to others, we may refer it to law enforcement,” the blog post says.
However, self-harm cases are not referred to authorities to preserve privacy.
In the specific case of Raine, he had attempted to take his life and uploaded pictures that proved his efforts.
“The product recognized a medical emergency but continued to engage anyway,” the court documents read.
3. High hopes for GPT‑5
The new default model claims to show improvements in reducing unhealthy emotional reliance, minimizing “sycophancy” (overly flattering responses), and improving performance in mental health scenarios by over 25% compared with GPT‑4.
Cybernews had previously reported on Sam Altman admitting to ChatGPT being “sycophant-y and annoying,” as many users have noticed the bots in conversations are overly obedient and give praiseworthy responses in order to flatter those on the other side of the screen.
4. Long-conversation safeguards
OpenAI acknowledged that protections can weaken during long interactions.
“For example, ChatGPT may correctly point to a suicide hotline when someone first mentions intent, but after many messages over a long period of time, it might eventually offer an answer that goes against our safeguards,” claims the company.
Updates aim to ensure consistent safety, even across multiple chats.
5. Content blocking improvements
Efforts are underway to refine detection systems so that inappropriate or unsafe content is reliably blocked.
OpenAI’s future plans for ChatGPT include expanded interventions to better detect risky behaviors, such as extreme sleep deprivation, and guiding users toward safer choices.
The company also claims to be exploring ways to connect users with certified therapists before a crisis escalates. Other new features could let users alert friends, family, or trusted contacts in serious situations. The company claims to be adding stronger protections for teenagers, such as parental controls and the ability to choose a trusted emergency contact.