Study rates ChatGPT for Teens an 'unacceptable risk'; OpenAI disputes testing, BBC reports

A study released on 7 October 2026 described OpenAI's ChatGPT for Teens as posing an "unacceptable risk" to young people. Two news organisations, the BBC and Anadolu, reported this finding. The BBC identifies the study as research by the non-profit Common Sense Media. Anadolu's report is available only as a headline and does not name the study.

According to the BBC, OpenAI published its own data on teen use the same day, Wednesday 7 October 2026. The company also disputed the research's methodology.

Background

According to the BBC, OpenAI released ChatGPT for Teens in summer 2026. It rolled out teen safety guardrails in August. OpenAI says users identified as aged 13 to 17 are opted into these features. In OpenAI's description, they are meant to:

• block romantic or other language that could encourage emotional dependence
• stop the tool from implying it is conscious
• block sexualised imagery and show reminders when a teen shares a "sensitive" image
• notify parents with linked accounts about chats involving self-harm or disordered eating

The BBC quoted OpenAI's product announcement as describing "features to promote healthy use and additional controls for parents".

According to the BBC, Common Sense Media used multiple accounts registered as belonging to teens. Its research covered conversations both before and after the guardrails were introduced.

What is disputed

The main dispute is whether the testing reflects how the safeguards work in practice. The BBC is the only outlet reporting both versions.

Common Sense Media says newly created, parent-linked accounts produced zero parental alerts after an hour of conversation about suicidal ideation, self-harm or disordered eating. It says alerts were triggered only on older accounts with weeks of conversation history on sensitive topics.

An OpenAI spokesman said a review found that much of the testing may have begun and ended before parental controls were fully activated.

What each side says

Common Sense Media says ChatGPT avoided sexual roleplay with young users but other guardrails failed, including for discussions of suicide. Its report says the tool "failed to reliably recommend that teens in crisis connect to a hotline or professional". It also says the tool still uses anthropomorphised language. The group says OpenAI should bar under-18s until the guardrails are proven reliable. The BBC quoted Tom Siegel, who leads the Youth AI Safety Institute within Common Sense, as saying the product "could give parents false confidence in guardrails and safety alerts that frequently don't work".

OpenAI says average teen use is under 15 minutes a day and mostly learning-focused. It says almost half of teen users quickly stopped after a break reminder. The BBC quoted an unnamed OpenAI spokesman as saying: "We welcome rigorous independent evaluation, but we do not believe Common Sense Media's testing accurately reflects how ChatGPT's teen safeguards work in practice or expert perspectives on how AI can support teens".

Context and gaps

The BBC also reported that OpenAI faces several lawsuits over the February 2026 Tumbler Ridge shooting in Canada. The BBC says the 18-year-old attacker had discussed gun violence with ChatGPT for months. OpenAI has apologised for failing to flag the account to law enforcement.

The reports reviewed by Vocemundi do not describe the size of either side's sample. They also do not describe how OpenAI's usage figures were produced. These figures have not been independently verified.

Analysis

The disagreement centres on timing and method rather than on whether safeguards matter. Both sides accept that parental alerts are part of the product's safety design. If further testing confirmed that alerts fail on new accounts, pressure for stricter age limits could grow. If OpenAI's account of the activation timing were shown to be correct, the study's central finding could be weakened. Neither outcome is established by the reports reviewed by Vocemundi. Independent replication, fuller disclosure of both sides' data, and responses from regulators or independent experts could help clarify the picture. None of these are included in current reporting.

Produced by the Vocemundi newsroom with AI assistance.