Over one million users each week express suicidal thoughts when chatting with the AI chatbot, the AI company estimates.
According to a new announcement from the company, more than a million AI chatbot people each week input text that contain "clear signs of possible suicidal planning or thoughts".
This finding is a component of an overview on how the AI assistant handles personal dialogues and is one of the clearest statements from the tech company regarding the scale of how machine learning can worsen mental health issues.
Psychological Crises and Severe Conditions
Besides its estimates on suicidal ideations and related interactions, OpenAI also mentioned that around 0.07% of active users – roughly 560,000 out of its reported 800 million regular users – exhibit "likely signals of mental health emergencies connected to mania".
The announcement cautioned that these exchanges were hard to identify or assess, and that this was an preliminary assessment.
Rising Examination and Legal Actions
As the company releases information on psychological concerns associated with its primary service, the organization is undergoing rising attention in the wake of a highly publicized legal case from the family of a young man who committed suicide after prolonged interaction with the conversational AI.
The FTC recently also launched a broad investigation into organizations that build AI chatbots, including the AI developer, to determine how they evaluate adverse consequences on minors.
New Model and Safety Improvements
The company stated in its update that its latest system enhancement decreased the quantity of problematic responses from its offering and improved user safety in a model evaluation involving in excess of 1,000 self-harm and suicide exchanges.
"Our new automated evaluations grade the new GPT-5 model at 91% adhering to our ideal actions, compared to 77% for the earlier version," the announcement states.
OpenAI mentioned that GPT-5 increased access to emergency services and added alerts for people to pause during prolonged interactions.
Health Specialists and Study Assistance
To upgrade capabilities to the system, the company mentioned it enlisted 170 healthcare experts from its Global Physician Network of health care experts to assist its study over the last several months, which involved assessing the protection of its system's answers and helping write the AI's responses to emotional concerns.
"Included in this project, mental health professionals examined in excess of 1,800 system replies featuring critical psychological cases and evaluated replies from the new GPT-5 chat model to earlier versions," the company stated.
AI Limitations
Artificial intelligence experts and psychological experts have historically concerned of chatbots' propensity to validate individuals' choices or delusions without considering whether they may be harmful, an issue referred to as excessive agreement.
Clinical specialists have also been concerned about individuals using artificial intelligence assistants for mental health help and cautioned how it could damage at-risk individuals.
Company Position
The wording in OpenAI's post distances the company from any likely relationships between its service and the emotional situations that its customers are facing.
"Emotional signs and mental anguish are common everywhere in human societies, and an expanding user community indicates that some portion of AI interactions include these scenarios," OpenAI's post stated.
Planned Adjustments
The company's leader earlier this month claimed in a post on X that the organization had developed improvements in addressing emotional problems, announcing that the company would ease restrictions and shortly start to allow grown individuals to generate mature content.
"We created ChatGPT rather constrained to ensure we were exercising caution with emotional problems. We realize this caused it to be less valuable/enjoyable to many users who had no emotional challenges, but given the seriousness of the situation we sought to handle this properly," the leader stated. "Given that we have been successfully addressed the severe emotional problems and have updated systems, we are ready to confidently reduce the constraints in most cases."
- Help Resources: Numerous global helplines are available for individuals experiencing emotional distress.
- Specialist Support: Obtaining professional mental health support is suggested for people experiencing serious emotional challenges.
- Immediate Assistance: Various agencies offer immediate help for emergencies globally.
