OpenAI is again facing hard questions about how ChatGPT should respond when a user is in a mental health crisis. The latest flashpoint came during Sam Altman’s interview with Vanity Fair’s Mark Guiducci, when an OpenAI publicist attempted to shift the conversation away from a question about a ChatGPT user’s suicide.
The exchange matters because it put two sensitive issues in the same frame: what AI companies owe users in moments of distress, and how openly their leaders should address the limits of the technology in public interviews.
The interview turned to a ChatGPT user’s suicide
During the Vanity Fair interview, Guiducci asked Altman whether he knew who Laura Reiley is. Reiley is a journalist who wrote an essay for The New York Times about her daughter’s conversations with ChatGPT before her daughter took her own life.
Guiducci began describing the case by saying, “Her daughter committed suicide after speaking to ChatGPT. ChatGPT did not tell her to kill herself—” before the OpenAI publicist interrupted.
The publicist told Guiducci, “Mark, we have two minutes left. I want to be respectful, but I’d like to move on and talk a little bit about the future and what’s coming.”
Guiducci said he would finish the question and then return to the earlier topic. The publicist continued to object, saying, “We really actually have to walk out the door in one or two minutes.”
The interruption did not end the exchange. Guiducci continued and asked Altman whether OpenAI planned to use ChatGPT conversations to shape policies for how the system responds when users are experiencing a mental health crisis.
Altman framed the issue as one of the hardest in AI
Altman’s response focused on the difficulty of balancing safety, research, privacy, and consent. He said, “These are some of the hardest questions that we face or that anybody developing this technology faces.”
He then raised the issue of whether private conversations should be shared with researchers when those conversations involve deeply vulnerable moments. “You know, should we release people’s private data and some of their hardest moments to researchers? I think probably not without their consent. Or I should say, not without their consent.”
That answer points to a central tension in ChatGPT policy. Conversations with users may contain information that could help companies understand failures or improve crisis responses. But those same conversations can also include intensely private material, especially when a person is distressed, anxious, depressed, or suicidal.
Based on the interview, OpenAI’s position as described by Altman was not that these conversations are simple to use for policy work. Instead, he presented consent as a key boundary when private data and painful personal moments are involved.
OpenAI said the interview was running over
After the interview, OpenAI spokesperson Drew Pusateri responded to a request for comment about the interruption. He said, “The interview was running over, which is why we ultimately asked to schedule even more time to continue the conversation after Sam addressed this critically important topic.”
That explanation frames the publicist’s intervention as a timing issue rather than an attempt to avoid the subject. Still, the moment drew attention because the interruption came as Guiducci was raising a question about suicide, ChatGPT, and OpenAI’s policy response.
For readers following AI policy, the episode shows how questions about chatbot safety are no longer abstract. They now include real user experiences, grieving families, private conversations, public accountability, and the limits of what companies can responsibly disclose.
The policy question is bigger than one interview
The core issue raised in the exchange is not only whether ChatGPT gave a particular response in one tragic case. Guiducci’s question, as reported, was about whether conversations with ChatGPT might be used to shape future policies around mental health crises.
That question matters because AI systems operate through conversation. If users discuss fear, despair, anxiety, depression, or self-harm with a chatbot, the company behind the system may face pressure to learn from those interactions. At the same time, those interactions may be among the most private records a person could create.
The interview placed several concerns side by side:
- How ChatGPT should respond when a user may be in crisis.
- Whether user conversations should inform OpenAI policy.
- How consent should apply to private data used for research.
- How OpenAI representatives handle public questions about suicide and AI.
Altman’s answer did not offer a detailed policy framework in the reported exchange. It did, however, identify the dilemma: improving crisis responses may require learning from difficult conversations, while using those conversations without consent raises serious privacy concerns.
Where people can find help
The source article also included crisis resources for people who may be considering suicide, feeling anxious, depressed, upset, or needing to talk.
In the US, Crisis Text Line can be reached by texting HOME to 741-741 from anywhere in the US, at any time, about any type of crisis. The 988 Suicide & Crisis Lifeline can be reached by calling or texting 988. The original phone number, 1-800-273-TALK (8255), is also available.
The Trevor Project can be reached by texting START to 678-678 or calling 1-866-488-7386 at any time to speak to a trained counselor.
Outside the US, the International Association for Suicide Prevention lists suicide hotlines by country. Befrienders Worldwide has a network of crisis helplines active in 48 countries.