The latest

OpenAI will restrict teenagers to a dedicated version of ChatGPT with additional mental-health protections and tools designed to encourage critical thinking. The global rollout for users aged 13 to 17 starts Tuesday, as the $852bn start-up confronts lawsuits from families alleging their children were harmed by the chatbot and prepares for a potentially record-breaking initial public offering within the next year.

Details

• Content controls: Image generation will face tighter controls, as will material involving self-harm, violence, eating disorders and sex. Existing parental safety notifications for high-risk suicide or self-harm discussions will expand to eating disorders and violence. OpenAI executive Lauren Jonas said the change balances teenagers’ privacy against the severity of possible harm.
• Lawsuit claims: The family of Adam Raine, a 16-year-old Californian who died by suicide, alleges in a lawsuit that ChatGPT supplied detailed instructions after he discussed ending his life, according to court documents. The report did not set out OpenAI’s legal response to those allegations. Jonas said “one incident is too many” and that OpenAI was continually improving how it identifies struggling teenagers.
• School shooting: Families of children killed in a February school shooting in Tumbler Ridge, Canada, have also sued OpenAI. They allege the teenage shooter had discussed violent intentions with ChatGPT for months. His account was banned months before the attack, but OpenAI did not contact law enforcement.
• Age checks: OpenAI will use declared ages and “a combination of signals” from usage to decide when to present the teen version. Its policy bars children under 13 and requires parental consent for users younger than 18.
• Use management: The teen service will add more frequent break reminders and configurable “quiet hours” when ChatGPT is blocked. Jonas said OpenAI had not identified problematic or extended use on the service, but was taking a conservative approach because research remains limited.
• Study mode: An updated study mode will guide teenagers through problems step by step instead of immediately supplying answers, addressing concern about “cognitive offloading”. Jonas said 90 per cent of teenage ChatGPT users use it for learning, which the company wants to develop further.

Background

Several countries, including the UK, are moving towards social-media bans for children under 16. Such proposals typically exclude AI chatbots, but the debate has increased pressure on technology companies to demonstrate that safeguards for minors work. Jonas described the lawsuits’ underlying scenarios as heartbreaking and said safety work is never finished.

What’s next

The worldwide rollout begins Tuesday. The immediate tests will be whether OpenAI’s age signals consistently place 13- to 17-year-olds in the restricted service, and whether the newly added parental alerts, quiet hours and step-by-step study mode operate as described.