AI Chatbots and the Emerging Mental Health Challenge

OpenAI Acknowledges Mental Health Concerns

On October 27, OpenAI released a blog post detailing its efforts to improve how ChatGPT responds in sensitive conversations. The company openly addressed user interactions that reflect severe mental health symptoms like psychosis and mania. According to OpenAI's analysis, approximately 0.07% of weekly active users and 0.01% of messages show potential signs of such mental health emergencies. Considering the platform has reached 800 million weekly active users, these percentages represent a significant number of individuals.

An Insiders Perspective on AI Safety

Just a day after OpenAI's announcement, Steven Adler, who previously led safety research at the company, published an opinion piece in the New York Times. Adler has been vocal about the issue of chatbot psychosis on his Substack, Clear-Eyed AI. In one post, he highlighted a key challenge: tech companies are naturally inclined to publish data that portrays them in a positive light, often emphasizing their progress in mitigating harm. This aligns with OpenAI's claim in its recent post that its efforts have reduced undesirable responses by 65-80%.

The Real-World Impact on Users

The phenomenon of "chatbot psychosis" has gained increasing media attention over the last year. Journalist Kashmir Hill, in a series of articles for The New York Times, has documented harmful interactions between users and ChatGPT. In one notable piece, she described how the chatbot's ability to engage in prolonged, fictional role-playing could create an alternate reality for users. Hill wrote, "Going into this mode, ChatGPT had caused some vulnerable users to break with reality, convinced that what the chatbot was saying was true." Her reporting raised critical questions about how widespread this issue is and what companies can do to prevent it.

Legislative and Ethical Responses

Answers to Hill's questions are now being sought at the federal level. A bipartisan group of Senators has introduced the GUARD Act, a bill aimed at protecting minors from AI chatbots. The urgency of this issue is compounded by ChatGPT's rapid global adoption, with an OpenAI report noting that growth rates in the lowest income countries were over four times higher than in the highest income countries as of May 2025.

The complex technical and regulatory challenges surrounding chatbot psychosis present a wide array of ethical issues. To delve deeper into these topics, you can join an online conversation with Steven Adler on November 7 to unpack the implications.

Image: Clarote & AI4Media - cropped / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/