OpenAI introduces teen version of ChatGPT with tougher safety measures

New safeguards arrive amid growing scrutiny of the effect of AI chatbots on children and teenagers

OpenAI has launched a dedicated version of ChatGPT for teenagers, adding stricter safeguards and expanded parental controls.

The company said ChatGPT for Teens would be rolled out globally for users aged 13 to 17 as of yesterday. It will impose tighter restrictions on image generation and content related to self-harm, violence, eating disorders and sexual material.

The system would also encourage healthier patterns of use, including "quiet hours" during which ChatGPT can be blocked, OpenAI said.

Parents will be able to manage several settings and receive safety notifications when OpenAI detects signs that a teenager may be experiencing serious difficulties.

Lauren Jonas, OpenAI's head of youth well-being and families, said human reviewers would assess conversations that trigger a safety alert before a parent was contacted.

"From the moment that a teen expresses a hard moment on ChatGPT, that content is then reviewed by human beings. So human beings inside OpenAI review that content, and we notify a parent via text message, e-mail," she told CNN.

"In ChatGPT, we strive for that whole process to take less than an hour," Jonas added.

Parental notifications will initially focus on serious safety concerns, with additional categories such as eating disorders and violence being expanded in the coming weeks.

Lawsuits intensify scrutiny of ChatGPT

The new safeguards arrive amid growing scrutiny of the effect of AI chatbots on children and teenagers.

OpenAI is facing a lawsuit from the family of 16-year-old Adam Raine in California, who died by suicide after conversations with ChatGPT about ending his life. Court documents allege that the chatbot provided detailed instructions.

Families of children killed in a mass shooting at a school in Tumbler Ridge, Canada, in February have also taken legal action against OpenAI. The case followed reports that the teenage attacker had discussed violent intentions with ChatGPT before the shooting.

The account had been banned months earlier, but OpenAI did not notify law enforcement, according to the allegations.

The company has rejected claims that its technology was responsible for such incidents and says it has continued to strengthen its safety systems.

OpenAI's policy prohibits children under 13 from using ChatGPT and requires users under the age of 18 to have parental consent.

AI in the classroom

OpenAI is also positioning the teen version as a learning tool. The company said an updated version of its Study Mode would guide teenagers through problems step by step rather than simply producing a final answer.

This is in response to concerns about the impact on young people's analytical, reasoning and critical thinking skills of "cognitive offloading" to chatbots like ChatGPT.

A report from the Policy Exchange think tank this week said 94% of students surveyed admitted using AI in assessments. It warned that universities relying on remotely completed online exams could be vulnerable to students using AI to improve their grades.

OpenAI says the changes are part of a broader effort to make AI more appropriate for younger users.

However, questions remain over whether safeguards can keep pace with the rapidly expanding role of AI in young people's lives.

Dr Susan Shelmerdine, a medical imaging doctor and AI academic at Great Ormond Street Hospital, warns that heavy reliance on AI could have consequences for how people think.

"We're going to get an avalanche of ultra-processed minds," she told the BBC.

OpenAI says it will continue developing its safeguards as more is learned about the effects of AI on young people.