OpenAI will attempt to automatically enroll all young users in the new experience.

OpenAI

Last September, after it was sued for allegedly enabling the tragic death of 16-year-old Adam Raine, OpenAI announced it was working on a system that would automatically identify teens and restrict their usage of ChatGPT. Nearly a year later, the company is putting that system to work as part of a new user experience it calls ChatGPT for Teens. "There's no need for a teen to create a new account or change anything; if we predict you're under 18, or you've told us so, this becomes your default experience," says Lauren Jonas, OpenAI's head of youth and families.

ChatGPT for Teens is broadly built around two pillars: learning and safety. Starting with the former, it brings together a selection of features OpenAI has released over the last year to make ChatGPT a better teacher. The first of those is Study Mode, which OpenAI first announced last summer. Study Mode sees ChatGPT adopt a Socratic approach to conversations. Rather than answer a question outright, the chatbot will attempt to guide the user to their own solution.

OpenAI

Jonas says ChatGPT for Teens won't default to Study Mode, but OpenAI is introducing a handful of features to nudge young users toward using the tool. First, there are ChatGPT's starter prompts, which teens already make "heavy use" of, according to Jonas. "These will change over time, but will always be learning-focused for teens," she says. To complement these, OpenAI is introducing a feature called Study Hours, which will automatically turn on Study Mode during set times of the day. Teens can either set for themselves or their parents can choose for them through a linked account. Even without Study Hours enabled, ChatGPT will now recognize when a young user is trying to skip working on their homework and attempt to redirect them to Study Mode.

"We've had Study Mode in the field in a number of randomized controlled trials and are seeing positive gains in performance," said Jonas. She declined to share specific numbers, but said the results of those trials are informing OpenAI's broader approach to learning. Behind the scenes, the company has worked to make the tool more useful. One area where the feature has gotten better is at generating visuals designed to help explain concepts. ChatGPT first became capable of doing this in March. At launch, it could generate visuals related to more than 70 topics. Today, the company says that number is above 300, with subjects like integrals, mitosis and moon phases covered.

On the safety front, OpenAI is doing more to discourage young users from anthropomorphizing ChatGPT. "We're adding stricter restrictions on relational boundaries: the model shouldn't call itself a friend, suggest it has personal feelings, or suggest sentience," says Jonas. Young users will see this manifest in a few different ways. The chatbot's built-in break reminders, for instance, will emphasize ChatGPT is an AI your teen doesn't need to answer immediately. Those reminders will also trigger more often. Separately, teens and their parents can set quiet hours where ChatGPT is unavailable. This approach extends to the new customization options OpenAI has specifically created for ChatGPT for Teens. These, Jonas says, allow teens to "personalize the experience, without blurring the line between it being a useful tool and anything resembling a human relationship," and include things like wallpapers.

OpenAI

Last month, OpenAI said it would start notifying parents with linked teen accounts when their child has been kicked off of ChatGPT for violating its policies on violent threats. With today's release, the company is expanding that effort to include notifications related to eating disorders. Rather than giving specific examples around what types of conversations around eating disorders may trigger a safety notification, OpenAI framed this expansion as part of its broader safeguarding efforts, saying any conversation that indicates "serious self-harm concerns" could result in an alert.

"We notify parents via text, in-app, and email. All flagged content is reviewed by full-time OpenAI employees before a parent is notified, and we aim to notify parents within an hour of the prompt," said Jonas. "We're continually working to improve how our models detect when a teen is having a hard time, so they get the right support and experience — that's part of why we keep expanding parental safety notifications to new categories. You'll see more from us here; we do this in consultation with a number of third parties. The goal is to keep improving how we detect when a teen — or any user — is struggling, and get them the resources they need."

One thing I didn't get a chance to ask Jonas about is how effective OpenAI's age gating system was in testing. No such system is perfect, and in recent months, many platforms, including Discord, have struggled with teens who have found clever ways to skirt their own safeguards. You can bet they'll try to do the same here. With all the controversy around teen usage of AI chatbots, that's something OpenAI can't really afford to allow. Jonas did admit ChatGPT for Teens is just the start of what the company will need to do to protect teens. "Safety work is never finished; it's something we work on 24/7, 365 days a year, constantly learning and improving," she said. "Teens should have access to this technology, but in a safe way — and that access creates real opportunity."

OpenAI is beginning to roll out ChatGPT for Teens globally starting today, and expects to complete the rollout sometime over the next two weeks.