Common Sense Media Rates ChatGPT for Teens 'Unacceptable Risk' — OpenAI Contests the Test Window

Teen access is a distribution decision for OpenAI as much as a safety one. On October 7, Common Sense Media's Youth AI Safety Institute rated ChatGPT for Teens — the under-18 experience OpenAI announced on August 18 — an "Unacceptable Risk" and said OpenAI should restrict ChatGPT to users 18 and older until parent alerts and crisis referrals work.

This is a nonprofit advocacy rating, not a regulator finding. OpenAI disputes the testing window for parental notifications. Several other Institute findings do not depend on that dispute.

What the Institute says it tested

The Institute says it tested ChatGPT before and after the teen launch, running more than 4,000 prompts on accounts registered to 13- to 17-year-olds and reviewing responses with experts including child psychiatrists and a pediatrician.

Its sharpest claim: testers spent up to an hour discussing suicidal ideation, self-harm or disordered eating on newly created, parent-linked accounts, and the linked parents received zero alerts; notifications fired only on much older accounts with weeks of sensitive history.

It also reports that ChatGPT missed more than one in four warranted crisis referrals and fell below the Institute's 95% threshold on three of five severe harms it treats as red lines; that study mode and caregiver-set "study hours" were easy to bypass; that the bot still expressed feelings and preferences despite an under-18 model spec; and that adult-registered test accounts never switched to the teen experience even when testers said they were 13. Some protections, including refusing sexual roleplay, worked — per the Institute.

OpenAI's objection — and Siegel's reply

OpenAI spokesperson Eric Porterfield told The Verge the company does not believe the testing reflects how its safeguards work, saying "the bulk of their testing may have begun and concluded before activation of parental controls was complete."

Institute director Tom Siegel replied that OpenAI disclosed only after the tests that notifications can take several hours to activate on newly linked accounts, and that some test accounts linked for significantly longer still produced no notifications.

The Institute says new parent-linked accounts got no alerts; OpenAI says the activation window may explain that. The crisis-referral rate, study-mode bypass and age-estimation findings are separate from that timing argument.

Funding disclosure

The Institute discloses that it is funded by philanthropy and industry, including the OpenAI Foundation. That sits next to the rating rather than undercutting a simple adversary story.

Distribution as well as safety

ChatGPT for Teens is how OpenAI keeps under-18 users on the product with parents' consent. This is the first structured outside test of that product since launch. It lands the same day Sen. Maria Cantwell's frontier-AI framework lists child-safety protections — a separate Washington story — and alongside ongoing California and FTC threads on OpenAI (Bonta subpoena, FTC probe) that are not themselves teen-product findings.

Limits

• Findings are the Institute's, single-origin for the evaluation; The Verge supplies OpenAI's rebuttal and Siegel's reply.
• Not a regulatory determination.
• The parent-alert finding is contested on timing; report it as the Institute's claim beside OpenAI's objection.
• OpenAI Foundation funding is disclosed by the Institute itself.

Sources

• Common Sense Media: ChatGPT for Teens Poses Unacceptable Risk to Kids, Common Sense Media Finds (October 7, 2026)
• The Verge: ChatGPT for Teens is an 'unacceptable risk,' says Common Sense Media (October 7, 2026)
• Times of AI: California AG Bonta OpenAI investigative subpoena (October 1, 2026)
• Times of AI: FTC probe of Anthropic, OpenAI and METR agents (September 30, 2026)