OpenAI is introducing a set of safety features for under-18s using ChatGPT, starting today. Teenagers with new accounts will be able to turn off the human-like voice response, reducing how much the tool appears to be a person. The company is also adding regular reminders for young users to take breaks, with messages noting that ChatGPT is AI and "it can wait."
OpenAI said these changes are not a reaction to any specific incident involving teens believing the chatbot is alive, but concerns are growing that people may become convinced AI tools are sentient. Last year, Microsoft's AI head Mustafa Suleyman said "seemingly conscious" AI products "kept him awake at night" due to societal impact worries.
The new teen accounts will become the default for users who identify as under 18 during sign-up, and for others OpenAI identifies as teens, though the company did not explain how it makes that identification. Under-13s remain barred from the product. The move follows a model similar to social networks' teen account protections.
Additional features include Study Mode hours, during which the tool will default to helping with schoolwork without giving away answers, and "quiet times" when the tool is switched off for the user. OpenAI reports that nine out of ten young ChatGPT users rely on it for learning.
OpenAI is also adding eating disorder-related prompts to a list of harmful requests that will be flagged to parents with linked accounts. Each alert will be reviewed by a human before being sent, with a goal of reaching parents within an hour of the request. The company says it will focus on "moments when offline support may matter most."
In the UK, there is no bespoke AI regulation yet, though the previous government under Sir Keir Starmer considered restricting AI chatbots for under-18s alongside social media reforms. Competing tools like Anthropic's Claude remain adults-only, and Google's Gemini also initially barred children. A recent report from the Policy Exchange think tank claimed most UK universities are exposed to AI-assisted grade inflation via remote online exams, with 94% of surveyed students admitting to using AI in assessments.
Dr Susan Shelmerdine, a medical imaging doctor at Great Ormond Street Hospital, previously told the BBC that doctors may one day ask patients about AI use like they ask about smoking or drinking, warning of "an avalanche of ultra-processed minds."
Source: BBC News