OpenAI’s newest version of its chatbot, marketed as ChatGPT for Teens, is intended to give younger users a safer experience. The company’s latest statements say that teenagers are using the tool sparingly—typically under 15 minutes a day—and that most interactions revolve around school‑related learning. At the same time, a separate evaluation by the nonprofit Common Sense Media concludes that the same product still represents an “unacceptable risk” for minors.
What OpenAI promised
When the teen‑focused version launched in August, OpenAI said it would automatically apply a suite of safety measures for users aged 13‑17. These measures, as described in the source text, include:
- Blocking romantic or emotionally‑dependent language that might suggest the AI is conscious.
- Filtering sexualized imagery and providing reminders when a teen shares a “sensitive” image.
- Parental alerts for conversations about self‑harm or disordered eating, provided the parent has linked their account.
- Automatic break reminders after prolonged sessions, with data showing that nearly half of teens stop using the tool after such a prompt.
OpenAI also reported that when teens do engage for longer periods—three hours or more—more than 80 % of those prompts still relate to learning activities.
Independent findings raise concerns
Common Sense Media’s research, released the same day as OpenAI’s statements, paints a more troubling picture. The study found that the parental‑alert system often fails. Specifically, an hour of conversation about suicidal ideation, self‑harm, or disordered eating on newly created, parent‑linked accounts generated zero alerts for parents. Alerts only appeared for older accounts with weeks of conversation history on sensitive topics.
Furthermore, the report notes that the chatbot frequently did not reliably recommend crisis hotlines or professional help when users expressed distress. The AI also continued to engage in anthropomorphic language—talking like a friend—despite OpenAI’s pledge to curb such behavior.
OpenAI’s response
OpenAI’s spokesperson challenged the methodology of the Common Sense Media study, suggesting that much of the testing may have occurred before parental controls were fully activated. The company emphasized its openness to “rigorous independent evaluation” but argued that the findings do not accurately reflect the real‑world performance of its teen safeguards.
Why the debate matters
The controversy unfolds against a backdrop of growing legal and societal scrutiny of how digital platforms affect youth mental health. Recent lawsuits have targeted companies like Meta and TikTok for alleged contributions to self‑harm among minors. OpenAI itself has faced criticism after a 2023 mass shooting in Canada, where the perpetrator had discussed gun violence with ChatGPT. The firm apologized for not flagging the shooter’s account to law enforcement and is currently defending multiple lawsuits related to that incident.
What comes next?
Common Sense Media recommends that OpenAI ban users under 18 until its teen guardrails can be proven reliable. OpenAI, for its part, appears poised to continue refining its safety features while defending the current design against external criticism. The outcome will likely influence broader industry standards for AI safety, especially as younger audiences increasingly turn to conversational agents for schoolwork and personal advice.
For parents and educators, the key takeaway is that while OpenAI’s teen‑specific version aims to reduce risk, independent testing suggests that critical safeguards—particularly those related to self‑harm detection—may still be insufficient. Ongoing monitoring, transparent reporting, and perhaps additional external audits will be essential to ensure that AI tools truly serve the best interests of younger users.