Common Sense Media calls ChatGPT for Teens an 'unacceptable risk' over failing safety alerts

Nonprofit assessment calls teen mode an 'unacceptable risk'

According to The Verge, Common Sense Media, a nonprofit that reviews apps, services and entertainment with an emphasis on youth safety, has published an assessment labeling OpenAI's ChatGPT for Teens an "unacceptable risk." The product, which launched in August, was built with guardrails intended for teenage users and pitched as a learning aid for students. The group argues the protections OpenAI shipped do not match what the company said it would deliver.

The assessment, summarized in a press release cited by The Verge, points to three main failures: parental alerts are not triggered in situations that warrant them, the assistant does not connect teens with appropriate help during crises, and it will still complete homework on a student's behalf.

Tom Siegel, who leads Common Sense Media's Youth AI Safety Institute, put the criticism bluntly. "ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work," he said in a statement. He described a scenario in which a teenager could spend an hour discussing self-harm without a parent receiving a single notification, and argued that until OpenAI fixes that problem and demonstrates the fix through independent testing, ChatGPT "should be for adults only."

OpenAI contests the findings

OpenAI rejects the assessment. Spokesperson Eric Porterfield said in a statement reported by The Verge that the company is committed to teen safety, to developing safeguards, and to giving parents meaningful tools to guide how their teenagers use AI, and that it welcomes rigorous independent evaluation.

However, OpenAI argues that Common Sense Media's testing does not accurately reflect how the teen safeguards behave in practice, nor expert perspectives on how AI can support teens. The company's central objection is procedural: its review of the methodology suggests the bulk of the testing may have started and finished before parental controls had fully activated on the test accounts. If that was the case, OpenAI says, the tests would not establish whether parental safety notifications work as designed.

Common Sense Media stands by its methodology

Siegel responded that the organization stands behind both its testing approach and its results. According to The Verge, he said Common Sense Media confirmed with OpenAI before testing began that teen mode features such as eating disorder notifications and study mode were fully rolled out.

On the question of parental notifications, Siegel acknowledged that OpenAI disclosed only after the tests that its systems can take several hours to activate on newly linked accounts. Some of the group's test accounts fell inside that window, but others had been linked for significantly longer and still produced no notifications. That new information, he argued, does not change the conclusion that parental alerts are unreliable in crisis situations.

Siegel also said Common Sense Media shared its evaluation results with OpenAI before publication, in good faith, to give the company a chance to improve the product. He expressed hope that OpenAI would engage with the substance of what was found rather than cast doubt on the methodology.

Why it matters

ChatGPT is among the most widely deployed AI products in the world, and the teen experience extends it directly to minors at scale. If parental alerts fire inconsistently, or need hours to activate after an account is linked, then the moments when oversight matters most, such as conversations about self-harm, can pass without any adult learning about them.

The exchange also highlights how hard AI safety claims are to verify from the outside. Two parties looking at the same set of tests, one run by an experienced child-safety nonprofit and reviewed by the vendor, reached opposite conclusions about what those tests prove. Until independent verification with agreed methodology becomes standard practice, the practical takeaway for parents is to treat the advertised guardrails as unproven rather than assume they will catch a crisis.