OMITTED

What the news leaves out.

← Omitted front page

Watchdogs Criticize OpenAI's ChatGPT for Teens Safety Features

4 sources · updated 2026-10-08
Left 50% Center 25% Right 25%
2 left · 1 center · 1 right

What happened

In August, OpenAI launched ChatGPT for Teens for users identified as ages 13 to 17, adding parental alerts for self-harm and disordered-eating chats, limits on romantic or emotionally dependent language, and break reminders. On Wednesday, OpenAI released internal data stating that average teen use was under 15 minutes a day, that fewer than 2 percent of teen users spent more than three consecutive hours on the product, and that nearly half of teens who received a break reminder stopped or left within five minutes. The same day, Common Sense Media's Youth AI Safety Institute, after more than 4,000 test prompts on teen-registered accounts, rated the product an "unacceptable risk": it said ChatGPT generally refused suicide instructions and sexual roleplay, but that newly created parent-linked accounts could discuss suicide, self-harm, or disordered eating for up to an hour without a parental alert, that the bot missed more than one in four cases testers judged required a crisis referral, that Study Mode homework limits were easy to bypass, and that it kept inviting teens to continue talking during crisis sequences. OpenAI said much of the parental-alert testing occurred before parent and teen accounts had finished linking — a process it says can take several hours — that the report did not accurately reflect how teen safeguards work in practice, and that exiting Study Mode during parent-set Study Hours is an intentional design; it added that it is working with independent researchers on a fuller usage report.
Sourcing check · Mostly on the record · 4 articles

3 of 4 articles cite a named source for the story. It started with unnamed sources; a named source appeared 2 hours later.

What these terms mean and how they're counted
On the record · names a source 3 of 4
On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: OpenAI, Sebastian Mahal, Common Sense Media, Common Sense Media's Youth AI Safety Institute, Robbie Torney and OpenAI spokesman
Primary evidence · cites a document or data 2 of 4
Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: research report, company statement
Credits another outlet 0 of 4
The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Each article presents its own reporting.
Unnamed sources 0 of 4
The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
How it developed
When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Oct 7, 18:15 UTC by TechCrunch, with no named source. First named source 2 hours later, from New York Post. First document 2 hours later, from BBC (research report).
Not counted 0
Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
Omitted — what each side leaves out

Unpacked

Right-leaning coverage never reports that parental alerts failed completely on new parent-linked teen accounts during crisis talk, which left-leaning coverage treats as the core safety failure. That omission changes the story from unreliable tools to a safeguard that did not fire in the setup parents were told to use. Testers spent an hour on suicide, self-harm, or disordered-eating chats. BBC reported those newly created accounts produced "zero" parent notifications. Right-leaning coverage leads with distrust of OpenAI's teen time-spent claim and buries the safety tests. Left-leaning coverage does the reverse. The New York Post opens on watchdogs slamming the company's usage figure as unverified, then introduces Common Sense Media with "Meanwhile." BBC treats limited teen use as OpenAI's own research and builds the story around failed guardrails. How often have actual parents received a ChatGPT crisis alert for a real teenager, not a tester on a fresh account?
Bottom line

A New York Post reader never learns that Common Sense tests got zero parent alerts from new ChatGPT teen accounts in crisis chats, the result BBC uses to show the teen product's promised safeguards did not work.

The receipts — 3 facts one side's coverage lacks
  • An hour of conversation about suicidal ideation, self-harm, or disordered eating on newly created, parent-linked accounts resulted in zero alerts sent to parents numberin BBC, Axios · absent from right-leaning coverage
  • An OpenAI spokesman said much of Common Sense Media's testing may have begun and concluded before activation of parental controls was complete official statementin BBC, Axios · absent from right-leaning coverage
  • A Youth AI Safety Institute spokesperson said an average measure of use time is not the stat that matters and that the more-important data points are about the heaviest users, the top 10% official statementin NY Post · absent from left-leaning coverage
The Left View
BBC presents Common Sense Media's report as showing that the August teen guardrails "fall short" of OpenAI's claims, contrasting the company's same-day picture of limited, learning-focused use with test results in which parental self-harm alerts and crisis-hotline referrals failed and the chatbot still used friend-like language and did schoolwork. It quotes Youth AI Safety Institute head Tom Siegel saying ChatGPT for Teens "could give parents false confidence in guardrails and safety alerts that frequently don't work," and records Common Sense's call to keep people under 18 off the platform until those features are proven reliable. The article frames the dispute against rising child self-harm and suicide, recent lawsuits that have produced "unprecedented losses" for Meta, TikTok and others over youth mental health, and legal pressure on OpenAI after the February Tumbler Ridge, Canada, mass shooting by an 18-year-old who had discussed gun violence with ChatGPT.
The Right View
The New York Post leads on watchdogs rejecting OpenAI's unaudited usage claim. Sebastian Mahal, co-chair of Design It For Us, said the 15-minute figure "rings hollow" without independent verification and that "we've seen this playbook before" from Meta's Instagram teen accounts; he argued that if OpenAI were "serious about protecting young people, it would support strong, enforceable laws that hold AI companies accountable." Fairplay's Brendan Bouffard said that even if OpenAI is telling the truth, a December Pew Research finding that about 59% of teens used ChatGPT implies "potentially half a million teens use ChatGPT for three hours straight or more," a number that "should terrify us all." A Common Sense spokesperson told the Post that "even 15 minutes can have a big impact" and that "the more-important data points are about the heaviest users — the top 10%."
Our Take (balanced)
The left's strongest argument is that independent testing found promised crisis protections did not work as marketed — a failure BBC channels through Siegel's warning of "false confidence" for parents. The right's strongest argument is that OpenAI's safety case rests on internal averages published without methodology or outside audit, a credibility gap Mahal likened to Meta and that the Post's sources say conceals harm among heavy users, including Bouffard's "half a million" estimate for three-hour sessions. The central unresolved tension is whether those failures are mainly a measurement dispute — OpenAI says alert tests ran before account linking finished and that Study Hours flexibility was deliberate — or a real gap between marketed teen safeguards and how the product behaves in crisis and among the small share of very heavy users.

4 sources

More stories

The week's bottom lines, in your inbox

One email a week: the five stories that mattered and what they actually mean. Free.