Soon, OpenAI says, parents will get more control over their children’s use of ChatGPT. In a blog post made public Tuesday, the artificial intelligence company explained how it aimed to have ChatGPT engage more actively and across a wider range of situations in which it notices possible mental health crises in consumers that might cause harm.
ChatGPT is finally receiving OpenAI’s new mental health safeguards after the report of first wrongful death case brought by two California parents, this declaration from the business follows closely the claim that their 16-year-old son’s suicide was caused by ChatGPT. Although OpenAI did not expressly reference the adolescent, Adam Raine, in its Tuesday update, the company hinted at upcoming changes following the lawsuit’s filing.
According to OpenAI, parents would be empowered to receive alerts of their children’s acute distress interactions with ChatGPT within the next month. The business aims to let parents link their accounts with those of their kids, set age-appropriate limits for ChatGPT’s answers, and control features such as the bot’s chat history and memory growth. Parents will also soon be able to get alerts when ChatGPT determines their adolescent is going through “a moment of acute distress,” as indicated in OpenAI’s blog article.
This would be the first capability that enables ChatGPT to flag a minor’s communications to an adult, a safety measure some parents have been pleading for. for fears that the chatbot could not properly de-escalate crisis events alone. Earlier this year, when Adam Raine shared his suicidal ideas with GPT-4o, the bot reportedly at times discouraged him from seeking the human help offered. As described in his family’s lawsuit, he helped in writing a suicide note and even advised on how to construct a noose.
While ChatGPT did repeatedly remind Adam of the suicide hotline number, his parents contend that their son readily avoided these messages. OpenAI admitted in a prior blog article following the announcement of Raine’s erroneous death suit that its present defenses are designed to guarantee ChatGPT provides sympathetic answers and points customers to actual resources.
