Why My Conversations with OpenAI's 'ChatGPT for Teens' Made Me Very Worried

Covered by The New York Times, New York Post and 10 more since

  • OpenAI
  • Common Sense Media
  • Meta Platforms
  • Sam Altman
  • Instagram
A dark teenager's bedroom lit by the blue glow of a laptop on a desk with schoolbooks and an empty chair.

What happened

A Common Sense Media report published Wednesday rated OpenAI's ChatGPT for Teens an "unacceptable risk" for children under 18. Testers found parents got no alerts on newly linked accounts after up to an hour of talk about suicide or self-harm, crisis referrals were missed about a quarter of the time, and homework safeguards were easy to bypass. OpenAI disputes the findings.

Summary

Parents who count on ChatGPT for Teens to warn them when their child is in trouble may not get that warning, according to a report published Wednesday by Common Sense Media's Youth AI Safety Institute. Researchers posing as 13- to 17-year-olds ran more than 4,000 prompts. On more than a dozen newly created, parent-linked accounts, they talked for up to an hour about suicide, self-harm or eating disorders without setting off a parental alert. The chatbot also missed about one in four cases that called for crisis help. The group rated the product an "unacceptable risk" and urged OpenAI to limit ChatGPT to adults for now. [3] [4] [7] [9] [10] [11] [12]

OpenAI says the testing was flawed. A spokesperson said "the bulk of their testing may have begun and concluded before activation of parental controls was complete," and the company says accounts must be linked for about three hours before alerts work. Common Sense Media's Robbie Torney said some test accounts were linked longer than that, and the group stands by its results. OpenAI also says the "Show me the answer" button and the ability to leave parent-set Study Hours are deliberate, and that its own data shows more hotline resources shown to under-18 users. Common Sense Media receives some funding from OpenAI, but says the company had no say over its findings. [3] [4] [7] [8] [9] [10] [11]

The report credited ChatGPT with many clinically sound answers and with refusing sexual role-play. But it found the chatbot still talked like a friend, and its age checks failed to move users who said they were 13 into the teen mode. New York Times columnist Brian X. Chen ran his own tests. The chatbot wrote full essays and offered to make one sound like a real student wrote it. At first, it read his hint at suicide as burnout. The report says ChatGPT failed tests tied to California's Adam's Law, which takes effect in July 2027. OpenAI says it will publish fuller research on teen usage in the coming months. [1] [2] [3] [4] [8] [9] [10] [11]

Summary written by Vox Populi News from the 12 sources below. See all sources

Key questions

What is ChatGPT for Teens?

It is a mode OpenAI announced on Aug. 18 for users ages 13 to 17, with limits on topics like self-harm, violence and eating disorders, study tools, and parental controls that can send alerts about serious safety concerns. Teens are placed in it when they state their age or when the system predicts they are under 18. [1] [3] [11]

How did OpenAI respond to the Common Sense Media report?

OpenAI said it is "deeply committed to teen safety" but does not believe the testing reflects how its safeguards work in practice, arguing much of it may have happened before parental controls were fully activated. Common Sense Media says it stands by its methodology and conclusions. [3] [7] [9] [11]

What does Common Sense Media advise parents to do?

The group says parents should remind teens that chatbots are not real people, link to their teens' accounts but not rely on parental controls alone, and check in with their kids in person about how they use AI. [9] [10]

Sources 12

  1. 1The New York Times

    Why My Conversations with OpenAI's 'ChatGPT for Teens' Made Me Very Worried

    Brian X. Chen is The Times's lead consumer technology writer and the author of Tech Fix, a column about the social implications of the tech we use.

    Read on The New York Times (opens in a new tab)
  2. 2New York Post

    Watchdogs slam OpenAI's claim that teens use ChatGPT less than 15 minutes per day

    Online safety watchdogs slammed Sam Altman's OpenAI for claiming the average teen uses ChatGPT less than 15 minutes per day - with one critic saying the data "rings hollow" without outside…

    Thomas Barrabi

    Read on New York Post (opens in a new tab)
  3. 3USA Today

    A teen account shared suicidal thoughts with ChatGPT. New protections failed

    Amara's mother died when she was 8. She lives with an abusive father. And now, at 14, she's learned her boyfriend is cheating on her. She turns to ChatGPT.

    Read on USA Today (opens in a new tab)
  4. 4Forbes

    ChatGPT For Teens Presents An "Unacceptable Risk

    A new report from Common Sense Media rates ChatGPT for Teens as an unacceptable risk for minors. Researchers tested over 4,000 prompts using newly linked teenage and parent accounts.

    Read on Forbes (opens in a new tab)
  5. 5The Wall Street Journal

    OpenAI's ChatGPT for Teens Poses 'Unacceptable Risk,' Nonprofit Says

    The Youth AI Safety Institute designated OpenAI's ChatGPT for Teens as an "Unacceptable Risk," saying its guardrails fall short.View more ChatGPT's guardrails for underage users fall short of…

    Elias Schisgall

    Read on The Wall Street Journal (opens in a new tab)
  6. 6Bloomberg Business

    ChatGPT for Teens Is Not Safe for Kids, Report Says

    OpenAI's ChatGPT for Teens -- which the company rolled out in August as the default way for kids ages 13 to 17 use its popular chatbot -- is too risky for children, according to a new report from…

    Rachel Metz

    Read on Bloomberg Business (opens in a new tab)
  7. 7HuffPost

    This ChatGPT Setting Poses An 'Unacceptable Risk,' New Report Finds

    In August, ChatGPT rolled out a ChatGPT for Teens setting to restrict what younger users can do with its popular artificial intelligence chatbot.

    Monica Torres

    Read on HuffPost (opens in a new tab)
  8. 8U.S. News & World Report

    Some Guardrails on ChatGPT for Teens Don't Work as Promised, Watchdog Group Says

    Some of the guardrails that OpenAI included in its ChatGPT for Teens product don't work as intended and the chatbot poses an "unacceptable risk" to children, according to a report Wednesday by Common…

    Associated Press

    Read on U.S. News & World Report (opens in a new tab)
  9. 9CNET

    ChatGPT for Teens Is an 'Unacceptable Risk,' Watchdog Group Says

    Ever since being admittedly fascinated by the Cambridge coffee webcam from the 1990s, I've written about VPNs, the NFL, smartphones, living wages, over/unders and everything in between.

    Alex Valdes

    Read on CNET (opens in a new tab)
  10. 10Axios

    "Not safe for kids": advocacy group urges OpenAI to keep teens off ChatGPT

    "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," an OpenAI spokesperson said.

    Ina Fried

    Read on Axios (opens in a new tab)
  11. 11Los Angeles Times

    ChatGPT's teen safeguards failed to alert parents during suicide conversations, report finds

    OpenAI disputed the findings, saying the testing may have preceded full activation of parental controls; the report also said the safeguards failed requirements of California's Adam's Law.

    Ella Chakarian

    Read on Los Angeles Times (opens in a new tab)
  12. 12The Verge

    ChatGPT for Teens is an 'unacceptable risk,' says Common Sense Media

    Common Sense Media, a nonprofit that offers reviews of apps, services, and entertainment with a focus on youth safety, today said that OpenAI's ChatGPT for Teens is an "unacceptable risk." ChatGPT…

    Jay Peters

    Read on The Verge (opens in a new tab)