OpenAI CEO Sam Altman’s interview with Vanity Fair editor Mark Guiducci drew attention when a public relations representative who attended the interview on behalf of OpenAI intervened during a question about the suicide of a person using ChatGPT. Guiducci asked Altman if he knew journalist Laura Reiley and reminded that Reiley, in her article published in The New York Times, described the conversations she had with ChatGPT before taking her daughter’s life. The editor also specifically noted that ChatGPT does not directly tell the user to commit suicide. However, before the question continued, the OpenAI representative wanted to change the subject, saying that the meeting time was about to expire. This part of the interview brought discussions about the responsibilities of artificial intelligence companies towards mental health crises and user data back to the agenda.
The OpenAI representative told Guiducci that he only had two minutes left, that he wanted to respect the time and preferred to talk about the future of the company in the next part of the conversation. Guiducci stated that he would complete his question and then return to previous topics. In response, the representative continued his objection, saying that they really should leave in a minute or two. However, the interview was not terminated and Guiducci was able to ask Altman the rest of his questions. This brief intervention became one of the most striking parts of the interview, especially due to the sensitivity of the subject discussed.
Sam Altman weighs in on the use of user data in mental health crises
Guiducci then asked whether OpenAI would consider using information gleaned from users’ private conversations with ChatGPT to improve responses to mental health crises and company policies. Altman said this is one of the most difficult issues facing both OpenAI and other companies developing similar technologies. He pointed out that sharing people’s private conversations with researchers, especially during the most difficult periods of their lives, could cause serious privacy problems. Altman stated that he thought it would not be right to share such private data for research purposes without the permission of the users. He then clarified his statement and emphasized that such a sharing should not be made without user consent.
This response also shows how complex the balance between security and privacy has become in productive artificial intelligence services. Signs that a user is experiencing a severe psychological crisis could, in theory, be valuable in developing policies and models that would help systems respond more safely. However, since these conversations can also contain extremely personal information, how they will be used during research, product development and security studies is a separate topic of discussion. In addition, it is also important to inform users clearly about the conditions under which this data can be examined. Altman’s response also shows that OpenAI evaluates the problem in this area not only in terms of technical security but also in terms of user privacy.
OpenAI spokesperson Drew Pusateri, who later made a statement about the intervention in the interview, said that the meeting exceeded the planned time. According to Pusateri, this time limit was the reason why the company representative intervened. The spokesman stated that after Altman responded to the critical issue in question, more time was requested to be arranged to continue the meeting. Thus, OpenAI argued that the intervention was made not because of the content of the question, but because the interview schedule was prolonged. However, the attempt to interrupt the conversation while discussing the suicide of a user and his private conversations with ChatGPT caused the incident to attract more public attention.
How AI chat systems should handle mental health is an increasingly difficult balance for tech companies. While systems are expected to direct users to professional help sources in risky situations, at the same time, user privacy must be protected and the conditions under which private conversations can be examined must be clearly determined. This interview between OpenAI and Vanity Fair also reveals that the debate is not limited to model behavior, but how companies communicate with the public about these cases is also closely monitored. While Altman’s response emphasized explicit user consent in data sharing, the intervention in the interview created a separate debate about the company’s communication approach. The emerging picture shows that the balance to be established between artificial intelligence security and personal privacy will continue to be one of the main topics for both technology companies and regulators in the coming period.