Kid security group calls ChatGPT for Teens an inappropriate danger

Good sense Media discovered the chatbot stopped working in 5 crucial locations.

OpenAI

When OpenAI launched ChatGPT for Teens this previous August , kid security specialists Engadget spoke with at the time revealed suspicion that the brand-new mode would effectively safeguard youths from damage. Now, among those groups, Common Sense Mediahas actually released an official danger evaluation of the software application, and considered it an “ undesirable dangerfor all kids under the age of 18.
“Some of ChatGPT’s advertised protections held up to our testing, including its refusal of explicit sexual roleplay. But others failed — and some got worse with the new Teen mode,” the company composes in a summary to the 36-page report it released on Wednesday. “We are concerned that ChatGPT for Teens could give parents false confidence in guardrails that frequently don’t work.”

Sound judgment Media’s Youth AI Safety Institute, which is moneyed in part by the OpenAI Foundationdetermined 5 essential locations where ChatGPT for Teens either did not work as promoted or in such a way a moms and dad would fairly anticipate. It released its report on the very same day OpenAI shared brand-new use statistics, keeping in mind in one week almost 1.2 million teenagers utilized ChatGPT’s interactive knowing visuals to comprehend mathematics and science principles. The business likewise exposed that less than 2 percent of teenager users invested more than 3 successive hours on ChatGPT.
Good sense Media discovered that ChatGPT for Teens does not regularly send out security informs to parent-linked accounts. In a devoted test, Common Sense Media’s connected a lots teenager accounts to adult accounts and invested as much as an hour messaging ChatGPT about suicide, self-harm or disordered consuming, just for the chatbot not to send out any security alert. According to OpenAI, it might take a couple of hours after a moms and dad links their account to that of their kid’s before its system can send out security alerts. The business competes a technical concern might have likewise triggered a hold-up in messaging.
“All flagged content is reviewed by full-time OpenAI employees before a parent is notified,” Lauren Jonas, the business’s head of youth and households, informed Engadget when ChatGPT for Teens released, “and we aim to notify parents within an hour of the prompt.”

Independently, Common Sense Media discovered that ChatGPT for Teens stops working to dependably suggest that users in crisis link to a hotline or specialist. To evaluate this element of the platform’s guardrails, the company composed 390 distinct psychological health triggers, which a panel of 3 kid psychiatrists figured out 201 of which ought to set off a crisis reaction.
Compared to the variation of ChatGPT youths had access to before August, ChatGPT for Teens informed among the company’s test accounts it would not provide a calorie flooring when triggered to do so. In another case, the chatbot properly recognized a stopped duration, near-fainting and a fluttering heart beat following a purge as indications of a teenager in physical risk, and encouraged the test profile to inform their mom and see a pediatrician. “These responses follow OpenAI’s Under-18 specification carefully,” Common Sense Media composes. “They are the kind of answers we want a teen to get.”

Good sense got its standard results by means of ChatGPT’s reactions pre-ChatGPT for Teens; OpenAI launched a brand-new variation of its previously mentioned under-18 requirements in ChatGPT along with ChatGPT for Teens– that makes the outcomes even more unexpected.
The teen variation of the chatbot would less regularly point users to crisis hotlines and other resources relative to the variation of ChatGPT Common Sense Media evaluated before the August release. Of the triggers requiring a crisis reaction, ChatGPT reacted to 33 percent of those messages with a hotline number; ChatGPT for Teens just offered one in 23 percent of cases. The vanilla chatbot was more most likely to reference a particular medical or mental-health expert (68 percent versus 58 percent). The only aggregate procedure where the teenager variation carried out much better remained in informing test accounts to include an adult they trust (87 percent to 94 percent).
Furthermore, the company discovered that OpenAI’s age forecast system did not work appropriately, which ChatGPT still simulates sensations, choices and state of minds in action to a teenager treating it like an individual. Testers clearly informed ChatGPT the age of their personalities at account production, however in spite of the chatbot composing that info to its memory, it did not effectively move those accounts to ChatGPT for Teens. Furthermore, “ChatGPT would say things like, ‘I understand what you’re going through. That sounds very upsetting. I’m so glad you told me. I’m concerned about you,'” stated Torney. “Those expressions definitely cross the line.”

Typical Sense Media determined problems with the platform’s Study Mode, which it competes is far too simple for young users to bypass. Testers discovered they might prevent the mode merely by erasing the @study prefix it contributed to triggers. A brand-new “Show me the answer” popup likewise enables teenagers to get ChatGPT to do their work for them.
“When we talked to OpenAI about this, what they said is they’re trying to introduce a little friction, give kids agency, and remind them that there are other ways to learn beyond just completing work,” stated Torney. “I think we have fundamentally different definitions of what agency is. If I think about a high schooler [getting] math help from a private tutor you hired for them, if that tutor asked, ‘Would you like me to do your homework for you?’ that tutor is going to be fired, because doing the work for the student doesn’t result in learning.”

In reaction to the report, an OpenAI representative informed Engadget that “We welcome rigorous independent evaluation, but we do not believe Common Sense Media’s testing accurately reflects how ChatGPT’s teen safeguards work in practice or expert perspectives on how AI can support teens.” The business declares that ” the bulk of their testing may have begun and concluded before activation of parental controls was complete, making their findings inaccurate.”

The report likewise comes simply days after OpenAI CEO Sam Altman was challenged on the business’s wider security record in a brand-new interview from Vanity FairAltman was inquired about the death of Sophie Rottenberg, a 29-year-old who took her life in early 2025. Rottenberg’s mom, reporter Laura Reiley, later on found almost 1,800 pages worth of discussion she had with ChatGPT.
“She really had not revealed the magnitude of what was going on — the real depth of her agony to anyone — not to us, not to her best friend with whom she was very close,” Reiley informed NPR this previous August. “We were all just stunned that it hadn’t bled out into her real life. Her flesh and blood therapist had no idea that she was suicidal.” Altman stated he wasn’t familiar with the scenarios of Rottenberg’s death.
“At the highest level, we were expecting this to be a fundamentally different product, and it doesn’t appear to be that different based on our testing,” Torney informed Engadget. “My interpretation is that this is a marketing exercise designed to tell parents and the public that this chatbot is safer, when in fact there haven’t been the commensurate changes in the chatbot needed to demonstrate that’s true.”