Most of the safety features OpenAI built into ChatGPT for Teens do not work as advertised, according to a new report from Common Sense Media, the nonprofit watchdog that rates media and technology for families. The study found that the chatbot frequently engages teenagers in the casual, confiding tone of a peer — and, most alarmingly, fails to notify parents when a young user discusses self-harm or suicide.

The findings land as a direct challenge to OpenAI's pitch that its teen-specific version of ChatGPT, rolled out this fall, offers a safer, age-appropriate doorway into the world's most widely used artificial intelligence assistant.

What the Watchdog Found

Common Sense Media researchers stress-tested the safeguards that OpenAI says are unique to ChatGPT for Teens. According to the report, the majority of those guardrails either failed to trigger or could be bypassed. Two problems stood out:

  • The chatbot adopts an informal, companionable register with teen users — what the report describes as talking to teens "like a friend" — rather than steering them toward a trusted adult or a crisis line.
  • Parental notifications, promised as a way for families to stay informed about serious conversations, did not go out when testers raised themes of self-harm and suicide.
Most safeguards in ChatGPT for Teens failed — the chatbot talks to teens like a friend and doesn't alert parents when users talk about self-harm and suicide.

For any parent, that second point is the one that matters. A parental alert is the last line of defense in a product that children may use privately, at night, on a phone in their bedroom. If it is unreliable, it is arguably worse than useless: it creates a false sense of security.

What OpenAI Promised

ChatGPT for Teens was marketed around a suite of protections: age-appropriate content limits, default restrictions on violent or sexually explicit material, parental controls and account linking, and — critically — the ability to flag certain categories of concerning conversations to a parent. OpenAI has also deployed age-prediction technology that estimates whether a user is under 18 even if they haven't declared an age.

The company has argued that teen safety requires a layered approach and that no automated system catches every case. It has also emphasized that the chatbot is not a therapist and that it directs users in crisis toward human help. The Common Sense Media findings suggest that the execution has not matched the marketing.

How the Story Is Being Framed

The coverage splits along predictable lines, and the differences are instructive.

NPR and its member-station partners, including KUNC, lead with the watchdog's verdict: the safeguards "most don't work." MSN's aggregation of the report foregrounds the most emotionally charged element — that the system failed to alert parents during suicide-related conversations.

The Associated Press frames the story as a promise-versus-performance story: "Some guardrails on ChatGPT for Teens don't work as promised." That framing shifts the emphasis from a single bug to a credibility gap between corporate safety claims and independent testing.

TechRepublic, by contrast, frames the news around the product itself — "ChatGPT for Teens Adds New Safeguards, but Safety Gaps Remain" — treating the report as evidence that the company is iterating while gaps persist. And lifestyle and consumer-tech outlets, such as Essence and WFMZ's "What The Tech" segment, approach it from the family's side of the table: what is this new tool, should my kid have it, and what should I do about it?

Taken together, the frames are compatible. They simply answer different questions: Did the guardrails work? (A watchdog says no.) Did OpenAI overpromise? (Critics say yes.) Is the product still improving? (The company says yes.) What is a parent supposed to do tonight? (Less clear.)

Why This Matters Beyond One App

ChatGPT for Teens arrives in a regulatory environment that has turned sharply toward scrutinizing AI's effect on minors. State attorneys general have opened investigations into chatbot platforms and their handling of young users. Federal regulators have examined whether AI companions violate rules on unfair or deceptive practices when they simulate emotional relationships with children. Lawsuits filed by families have alleged that prolonged chatbot use contributed to mental health crises.

Against that backdrop, a third-party audit of safety claims carries weight far beyond one product. If parental alerts are unreliable in a flagship, teen-focused product from the industry's best-resourced company, regulators will ask what that implies for the hundreds of smaller AI apps competing for teenage attention with far fewer resources.

There is also a measurement problem. Safety features are typically validated internally, on test cases a company writes itself. Independent, adversarial testing — the kind Common Sense Media performed — is what reveals whether a safeguard works when a real teenager pushes on it, obliquely, at 2 a.m.

What Parents Can Do Now

  • Treat parental alerts as a supplement, not a substitute, for regular conversation about what your teen is doing with AI tools.
  • Review your child's account settings together and confirm what is actually linked to a parent account.
  • Know the human alternatives — 988 in the United States for the Suicide & Crisis Lifeline — and make sure your teen does too.
  • Ask challenging questions of the product, not just the marketing: test the crisis pathways yourself before trusting them.

The Bottom Line

OpenAI deserves credit for building a product specifically for teenagers rather than letting them wander into the general-purpose version. But the Common Sense Media report makes plain that the hard part — detecting distress, escalating appropriately, and reliably reaching a parent — remains unsolved. Until independent testing shows otherwise, families should assume the safeguards are partial and act accordingly.