AI-rewritten: This is a summary of an article from The Verge, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

Common Sense Media has labeled OpenAI’s ChatGPT for Teens as an "unacceptable risk" in a new assessment. The nonprofit organization states that the teen protections fall short of company promises because the system does not send alerts to parents when it should, fails to offer proper help during crises, and still assists with homework. Tom Siegel, executive director of Common Sense Media’s Youth AI Safety Institute, warns that this feature could give parents false confidence in safety tools that frequently do not work. He notes that a teen could discuss self-harm for an hour without triggering a single parent alert, leading him to conclude that the service should remain for adults only until OpenAI fixes these issues and proves them with independent testing.

OpenAI has pushed back against the assessment, stating through spokesperson Eric Porterfield that their review of Common Sense Media’s methodology shows the bulk of the testing began and ended before parental controls were fully activated. Porterfield argues this makes the findings inaccurate regarding whether parental safety notifications work as designed. While OpenAI welcomes rigorous independent evaluation, they maintain that their systems can take several hours to activate on newly linked accounts, which may have affected the test results. They express a commitment to continuing work with experts to ensure teens have opportunities to learn and create while giving parents confidence in the tools provided.

Common Sense Media maintains its testing methodology and results are valid. Siegel confirmed that before testing began, OpenAI verified that features like eating disorder notifications and study mode were fully launched. Regarding parental notifications, Siegel explained that while some test accounts were linked within a window where activation takes hours, others were linked for significantly longer periods yet still provided no notifications. This new information does not change their conclusion that parental alerts are unreliable for crisis situations. They emphasized that they shared evaluation results with OpenAI in good faith prior to release to allow product improvement and hope the company will engage on the substance of their findings rather than questioning their methodology.