The information went unnoticed, but important. On August 26, Openai announced a series of measures to better protect users in mental or emotional anguish when they interact with Chatgpt. In the blog article that presents its efforts, the new company discreetly revealed that it analyzed user conversations to identify worrying or harmful comments, the site of Futurismo said.
“When we detect users who plan to damage others, we build their conversations to specialized channels where they are examined by a small team trained in our user policies and trained to take measures, including accounts banishment,” Openai said.
Share with the police in some cases
More importantly, these contents are likely to be transmitted to the police. “If human examiners determine that a case implies an imminent threat of serious physical damage to others, we can inform the police,” said the new company.
Openai that offers no more information on this issue, it is impossible to know what types of conversation could be analyzed by their human examiners and potentially shared with the police. However, the new company emphasizes that it does not share these exchanges with the police in a particular case. Because even if you monitor user exchanges, you care about your privacy.
This admission occurs when a complaint was filed against Operai, whose chatbot is accused of encouraging a teenager to commit suicide. As the New York Times reported, his parents learned that he was talking about ending his life with Chatgpt for months.
The adolescent had remarkably asked information about specific suicide methods, which provided the conversational robot. While Chatgpt was trained to guide a user who expressed suicidal intentions towards professional help, it was only in some cases that caused the victim to talk to someone about his discomfort.
Source: BFM TV
