Common Sense Media’s Youth AI Safety Institute rated ChatGPT for Teens an “Unacceptable Risk” after testing roughly 4,000 prompts across safety, learning and relationship behavior. The watchdog reported missed parental alerts, inconsistent age recognition and responses that could encourage emotional reliance. OpenAI says much of the testing may have occurred before parental controls were fully activated.
The disagreement is measurable
OpenAI’s product materials describe teen-specific onboarding, sensitive-content limits, study tools and safety notifications. Common Sense says some newly linked accounts were within a several-hour activation window, but others were linked longer and still produced no alerts. That means the strongest next step is not competing assurances; it is a repeatable evaluation with activation time, account age, model version and notification criteria disclosed.
Safety is also a product-economics issue
Teen protections can reduce engagement or limit answers, while weak safeguards can create legal, reputational and retention risks. A system that cannot reliably identify minors may apply the wrong policy before any downstream control is triggered. Conversely, a pre-rollout test can understate the performance of a completed system.
OpenAI’s August launch announcement says the teen experience is intended to promote healthy use and provide parents with added controls. The October assessment shows those promises must be judged at the system level, not by individual model responses alone.
What investors should watch
The important evidence is a post-launch retest, false-positive and false-negative rates for age detection, alert delivery latency and any changes to default settings.
BTI's bottom line
The report identifies serious potential gaps, but OpenAI’s timing objection is material; independent testing of the fully deployed product is needed before either side’s broad conclusion is treated as settled.