OpenAI is rolling out new safety features today for under-18s who use its AI chatbot, ChatGPT. For new teen accounts, the ability for the chatbot to respond using a human voice will be removed to reduce its potential to appear like a person.
The company is also introducing regular reminders for young users to take breaks from the tool, emphasising that ChatGPT is AI and "it can wait." OpenAI stated this is not a response to specific incidents of children believing ChatGPT is alive, but rather addresses broader concerns about people becoming convinced AI tools are sentient.
Teen accounts will become the default for users who identify as teenagers upon signing up, or those identified as under 18 by the firm. OpenAI has not detailed how it will identify these users. Under-13s are not permitted to use the product.
Further features include the ability for young people to set Study Mode hours, where the chatbot assists with schoolwork without providing answers, and to activate "quiet times" when the tool is switched off. OpenAI also plans to flag eating disorder-themed prompts to parents with linked accounts, with each alert reviewed by a human before being sent.