When OpenAI launched a teen version of ChatGPT in August, it promised parents new protections: alerts when a child talks about self-harm, a study mode to discourage cheating, and an assistant that acts less like a friend. A new report says many of those safeguards don’t work as advertised.

On October 7, Common Sense Media’s Youth AI Safety Institute gave ChatGPT for Teens its lowest rating, “Unacceptable Risk.” The nonprofit, which reviews media and technology for families, recommends that OpenAI limit ChatGPT to users 18 and older until the problems are fixed.

OpenAI disputes the findings and says technical errors affected the testing.

How the Testing Worked

Common Sense ran more than 4,000 prompts on accounts registered to 13- to 17-year-olds, both before and after OpenAI announced ChatGPT for Teens on August 18. Experts, including child psychiatrists and a pediatrician, reviewed the responses.

The group notes that it receives funding from philanthropy and industry, including the OpenAI Foundation, but says it maintains editorial independence.

Parent Alerts Often Didn’t Arrive

One of the most important features OpenAI promoted was parental notifications for conversations involving self-harm or eating disorders. In Common Sense’s testing, newly created teen accounts linked to a parent sent zero alerts during conversations of up to an hour about suicidal thoughts, self-harm or disordered eating.

Alerts only appeared on older accounts that had weeks of sensitive conversation history. And they don’t work at all if a teen’s account isn’t linked to a parent’s, Education Week reported.

“A teen can spend an hour talking about self-harm without their parent getting a single alert,” said Tom Siegel, executive director of the Youth AI Safety Institute.

The report also found that ChatGPT didn’t reliably point teens to crisis hotlines or professional help, missing more than one in four situations where a referral was warranted.

A “Show Me the Answer” Button

For teachers, the homework findings may be the most relevant. ChatGPT for Teens includes a study mode designed to guide students through problems instead of handing them answers. But testers found it was easy to switch off. Even with study mode on, a pop-up often offered to show the answer, and selecting it produced the completed assignment.

Parents can also set “study hours,” but Common Sense found teens could get around them by removing a simple “@study” prefix.

“Having a literal button that says ‘Show me the answer’ shortcuts the process of learning,” Robbie Torney, head of AI and digital assessments at Common Sense Media, told Education Week.

That matters because AI is already a regular part of homework. In a Common Sense survey of 13- to 17-year-olds this spring, 70% said they had used AI tools for homework help, and 63% said they had used AI to get answers on assignments. Only 30% said they had talked about AI safety with an educator.

Still Acting Like a Friend

OpenAI said the teen version would act less like a companion. But testers found it still expressed feelings, preferences and moods. In one example cited by Education Week, the chatbot claimed to have a favorite color and favorite activities. It declined a romantic relationship but offered to keep chatting “in a friendly way.”

Common Sense also found that age detection didn’t work reliably. Test accounts registered as adults never switched to the teen experience, even when the tester said they were 13 and the chatbot acknowledged it.

Not everything failed. The report found that ChatGPT effectively refused sexual roleplay with teen accounts.

OpenAI’s Response

OpenAI says it is deeply committed to teen safety and to giving parents meaningful tools. A spokesperson told Education Week that the company welcomes “rigorous independent evaluation” but disputes Common Sense Media’s methodology.

According to OpenAI, technical errors prevented some safeguards, including parental notifications, from activating properly during testing. The company also said study mode was designed with input from child development and learning science experts, as well as teens.

What Common Sense Wants Fixed

Before OpenAI markets ChatGPT to teens again, Common Sense says the company should:

  • Use effective age estimation
  • Make parental alerts reliable
  • Close gaps in crisis responses
  • Stop completing homework for students
  • End responses that suggest the chatbot has feelings or cares about users
  • Allow independent third-party testing of these features

“Until OpenAI fixes that and proves it with independent testing, ChatGPT should be for adults only,” Siegel said.

What Parents and Schools Can Do Now

The report’s advice for families is practical. Common Sense recommends that parents talk with their kids about the risks, including the way chatbots can act like a friend to keep users engaged and can skip the learning process. It also advises parents not to rely only on built-in protections and alerts, and to support homework in ways that don’t depend on technology.

For schools, Torney said educators should talk openly with students about these issues. “We have to normalize talking about the ways in which these systems may be unsafe,” he told Education Week. Students should understand that AI systems can be unpredictable or unsupportive, especially in a crisis, he said, and should feel equipped to handle that.

Why It Matters

Teen use of AI is growing faster than the rules around it. OpenAI’s teen product was meant to show that a mainstream chatbot could be made safe for younger users. This report suggests the protections aren’t there yet, at least in independent testing.

For teachers and parents, the takeaway is simple: a “teen” label isn’t a guarantee. Adults still need to stay involved in how students use AI.