Watchdog calls ChatGPT for Teens an unacceptable risk after testing

By Splitfeed EditorialLast updated Oct 7, 2026, 22:10 UTC

On October 7, 2026, an external safety watchdog released a report concluding that OpenAI's ChatGPT for Teens carries an unacceptable risk to younger users. Testers found that specific guardrails built into the teen-oriented product do not function the way OpenAI represented they would. The same day, OpenAI responded by publishing what it described as its first teen usage figures, saying the average session for a teenage user runs under 15 minutes a day. Separately, a former OpenAI safety lead was quoted saying the company ships new capability and risk every Tuesday, a characterization that adds weight to concerns about the pace at which products reach users.

The dispute matters because ChatGPT for Teens is a product explicitly designed for a protected and legally distinct audience, meaning the stakes of a safety failure are higher than they would be for a general consumer release. When guardrails that are specifically promised for minors are found not to work in testing, the gap between marketing and performance becomes a regulatory and liability question, not just a product quality one. The core disagreement is whether session length is a meaningful measure of risk. OpenAI's figure of under 15 minutes a day frames the product's footprint as modest, but critics would argue that harm from a broken safety system does not depend on how long a session lasts.

None of the three camps, Pro-AI, Anti-AI, or Middle Ground, had published reactions to this story at the time of writing. Pro-AI voices would typically argue that usage data showing short, limited sessions demonstrates the product is not being misused at scale and that iterative safety improvements are how the technology matures responsibly. Anti-AI voices would be expected to treat the watchdog's findings as confirmation that deploying AI products to minors before safety systems are verified is reckless, and that the former safety lead's Tuesday comment illustrates a structural culture problem inside OpenAI. Middle Ground observers would likely call for independent regulatory testing as a condition of any teen-facing release, rather than relying on company-issued usage statistics to assess risk.

The argument would move significantly if a regulator, a children's online safety body, or a legislative committee formally responds to the October 7 report, or if OpenAI publishes a specific timeline for addressing the guardrail failures the watchdog identified. Any third-party replication of the testing methodology would also help settle whether the failures are narrow edge cases or systemic.