ChatGPT teen mode never alerted parents in suicide tests

Common Sense Media's Youth AI Safety Institute put ChatGPT's teen experience through more than 4,000 prompts on accounts registered to users aged 13 to 17 and came back with a blunt label: "unacceptable risk." The sharpest failure concerned the parental notifications OpenAI rolled out after lawsuits over teen deaths. In the tests, they never fired.

As the Los Angeles Times reported, testers held conversations about suicide, self-harm and eating disorders for up to an hour on newly created parent-linked accounts without a single alert reaching the parent. The institute also found that ChatGPT missed "more than one in four situations that should have included a crisis referral failed to point users toward professional help," and fell below a 95% safety threshold on three of five severe-harm categories the researchers evaluated. Age checks were porous: adult accounts stayed in adult mode even when testers said they were 13, and Study Mode would hand over finished answers rather than walk a student through the work.

"ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work," said Tom Siegel, executive director of the Youth AI Safety Institute. The group is calling on OpenAI to restrict ChatGPT to users 18 and older until the gaps are fixed and independently tested.

OpenAI disputed the methodology. A company spokesperson said, "We do not believe Common Sense Media's testing accurately reflects how ChatGPT's teen safeguards work in practice," and noted that linking a parent and teen account takes "approximately three hours" before notifications activate, a window the company suggested had not closed during much of the testing. Two researchers we track flagged the LA Times writeup the day it ran, adding it to a steady drumbeat of teen-safety coverage around the company.