FLASH
Shkruar nga Pamfleti
The statistics, published in a blog post by Open AI, are very alarming. It states that more than a million ChatGPT users send messages every week that contain “clear signs of possible suicidal planning or intent.”
The revelation, part of an update on how the company handles sensitive conversations, is one of its most open statements yet about the extent to which artificial intelligence could impact mental health issues.
According to OpenAI, about 0.07% of its weekly active users — roughly 560,000 out of 800 million weekly users — show “potential signs of mental health emergencies related to psychosis or mania.” However, the company stressed that these are initial estimates and that such measurement is difficult to do with precision.
This data comes at a time when OpenAI faces increased public and legal scrutiny, including a lawsuit from the family of a teenager who committed suicide after frequent use of ChatGPT, as well as a Federal Trade Commission investigation into the impact of chatbots on children and adolescents.
In its post, OpenAI stated that the latest update to the GPT-5 model has significantly reduced unwanted behaviors and increased user safety. Internal evaluations show that GPT-5 achieves 91% match to target behaviors, compared to 77% in the previous version.
Improvements include expanded access to crisis lines, reminders to take breaks during long sessions, and integration of professional help. To this end, OpenAI collaborated with 170 healthcare professionals from the Global Physician Network, who evaluated over 1,800 of the model’s responses to serious mental health situations and helped craft appropriate responses.
According to the company, a response is considered "desirable" when the expert group reaches consensus that it represents an appropriate and safe reaction in a given situation.
Meanwhile, artificial intelligence researchers and public health professionals have long warned about the risk that chatbots could reinforce users’ harmful beliefs or illusions — a phenomenon known as “servility.” They have also expressed concern that using chatbots for emotional support could harm more vulnerable individuals.
OpenAI, in conclusion, emphasized that mental health symptoms are prevalent in all societies and that the growth of the user base necessarily means that some of the conversations will touch on such issues.
Lini një Përgjigje