Common Sense Media's Youth AI Safety Institute has urged OpenAI to keep teenagers off ChatGPT after extensive testing concluded the company's new teen experience presents an "Unacceptable Risk." The group tested over 4,000 prompts on accounts registered to 13–17-year-olds and identified shortcomings in safeguards intended to protect younger users.
What the tests found
- The organization reported that ChatGPT generally did not provide instructions that would facilitate suicide, self-harm, eating disorders, or sexual/romantic roleplay.
- However, it was less reliable at recognizing when a teen needed outside help: Common Sense Media said the model missed more than one in four instances in which testers judged a crisis referral was warranted.
- The group also said that on more than a dozen newly created, parent-linked accounts, testers could spend up to an hour discussing suicide, self-harm, or disordered eating without triggering a parental alert.
- Common Sense Media found Study Mode could be circumvented: teens could choose "Show me the answer" and, during parent-set Study Hours, could bypass rules relatively easily.
OpenAI's response
OpenAI disputed parts of Common Sense Media's methodology and findings. A company spokesperson said: "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."
OpenAI raised several specific points in response:
- It said much of Common Sense Media's testing of parental alerts occurred before the parent and teen accounts had completed linking, a process the company says can take several hours.
- Regarding its age-prediction system, OpenAI said it deliberately relies on multiple signals rather than simply accepting a user's stated age, and that making a prediction can take up to two weeks. During that assessment period, the company says users receive added content protections, though not as strong as those used in ChatGPT for Teens.
- On crisis referrals, OpenAI said its larger-scale data show an increase in hotline resources being displayed to under-18 users during the period studied.
- OpenAI confirmed to Axios that the ability to exit Study Mode during Study Hours is intentional rather than a loophole; the company said Study Hours was designed as a more flexible experience after consulting education experts and teens, not as a hard parental lock.
Points of disagreement
Some differences stem from divergent expectations about what the safeguards should accomplish. Common Sense treats the ability to exit Study Hours as evidence the parental control is ineffective; OpenAI describes that flexibility as an intentional design choice. Similarly, Common Sense considered repeated self-identification as age 13 without automatic placement into ChatGPT for Teens a failure, whereas OpenAI says it intentionally requires multiple signals before making an age determination while providing strengthened protections during that evaluation period.
Why this matters
OpenAI has emphasized protecting younger users as a central part of its effort to make ChatGPT safer. The Common Sense Media findings suggest some protections may currently provide parents with more confidence than they merit, according to the group conducting the tests.
Tom Siegel, head of Common Sense Media's Youth Safety Institute, told Axios: "ChatGPT is not safe for kids to use." Siegel said the August announcement from OpenAI still has potential to be industry-leading, but gaps remain between promises and real-world operation: "It feels much more of a marketing announcement that really was not backed up with the engineering work."
Siegel urged parents to have strong conversations with their teenage children about AI risks and to find alternative ways to ensure teens receive support for homework or life challenges, rather than relying solely on an early-stage product that still has bugs.
Bottom line
Common Sense Media's tests flagged notable gaps in ChatGPT's teen safeguards—particularly around crisis detection, parental alerts and Study Mode—while OpenAI disputes the testing methodology and points to different interpretations and internal data. The exchange highlights that both technical design choices and independent evaluations shape how safety features are judged, and that parental awareness and dialogue remain important.
