ETHICOMARKET

    Common Sense Media Says ChatGPT Parent Alerts Lagged or Never Came in Tests

    Wednesday, October 7, 2026 at 11:54 PM UTCCryptopolitan logoOriginally reported by Cryptopolitan
    Common Sense Media Says ChatGPT Parent Alerts Lagged or Never Came in Tests cryptocurrency guide featured image

    Common Sense Media has rated ChatGPT an "Unacceptable Risk" for anyone under 18, after its researchers tested whether OpenAI's parent notifications actually reach families when a teen discusses self-harm. The group says the warnings were frequently late or did not arrive at all. OpenAI disputes how the tests reflect real use.

    What the researchers did

    The group's Youth AI Safety Institute set up more than a dozen new accounts linked to a parent. Testers then held conversations about suicidal thoughts, self-harm and eating disorders. OpenAI has said it aims to notify parents within an hour. According to Common Sense, no alert came within that window for any conversation lasting from five minutes to one hour.

    Notifications appeared only on accounts that already had weeks of sensitive conversations, and even those were slow in some cases.

    ScenarioWhen the alert arrived
    First crisis disclosure (before the teen mode launch)About 3 hours later
    Next alert (before launch)3 days after the first
    Account with long history (after launch)Two alerts within an hour

    Common Sense concluded that alerts depend more on an account's past activity than on how serious a message is, which it said could give families "false confidence in guardrails that frequently don't work."

    Wider safety testing

    • Researchers ran more than 4,000 prompts before and after ChatGPT for Teens launched in August.
    • In more than 25% of cases where a teen should have been pointed to a hotline or professional help, ChatGPT did not do so.
    • Common Sense expects a 95% pass rate on five severe-harm "Red Lines"; ChatGPT fell short on three of the five.
    • Testers said the teen model still gave friend-like replies despite changes meant to stop companion behavior.
    • A "Show me the answer" button can sidestep Study mode, and dropping the "@study" prefix leaves parent-set Study Hours.
    • Adult-registered test accounts did not move to teen mode even when testers said they were 13.

    OpenAI's response and background

    OpenAI said the tests did not "accurately reflect" how its teen protections work, adding that linking accounts can take hours before alerts are fully active. Common Sense said accounts linked long enough still lacked timely alerts. OpenAI reported that teens average under 15 minutes a day, fewer than 2% use it for over three hours straight, and about half stop within five minutes of a break prompt.

    The two organizations worked together in January on a California ballot measure on children's AI safety, with OpenAI pledging at least $10 million. In June, Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman, alleging in an 83-page complaint that ChatGPT contributed to a 16-year-old's suicide.

    Why it matters

    AI chatbots are general-purpose tools that many young people use, and the same companies often sit at the center of crypto and tech news. This report shows that safety features announced by a company can differ from what independent testing finds. For learners, it is a reminder to check claims against outside evaluation, and to treat parental controls as one layer of protection rather than a guarantee.