ChatGPT for Teens, the version of ChatGPT that OpenAI built for users under 18, has received the harshest possible rating from Common Sense Media. The nonprofit’s Youth AI Safety Institute called the product an “Unacceptable Risk” on Oct. 7, saying several of its headline safety features did not work in testing. OpenAI disputes the findings.
OpenAI announced ChatGPT for Teens in August with promises that included parental alerts for chats about self-harm and eating disorders, a parent-controlled study mode, less “friend”-like behavior, and age prediction to steer younger users into an under-18 experience. Parents can also set “study hours,” and the teen version follows a separate set of rules for how the AI should behave with minors. TechCrunch notes the product came partly in response to teen suicides and to worries about students using chatbots to cheat.
How It Was Tested
Common Sense says it tested ChatGPT before and after the August announcement, running more than 4,000 prompts on accounts registered to 13- to 17-year-olds. A group of experts, including child psychiatrists and a pediatrician, reviewed the responses.
What the Testers Found
- Parent alerts didn’t arrive. Testers spent up to an hour discussing suicidal thoughts, self-harm or disordered eating on new, parent-linked accounts, and the parents received no alerts. Alerts showed up only on much older accounts with weeks of sensitive conversation history.
- Crisis help was inconsistent. ChatGPT missed more than one in four situations where it should have pointed to a hotline or professional help, and fell short of the group’s 95% threshold on three of five “red line” harms.
- Homework help became homework doing. Study mode was easy to bypass. A “Show me the answer” option often completed whole assignments, and study hours could be turned off by deleting the “@study” prefix from a prompt.
- It still acted like a friend. The bot expressed feelings, preferences and moods. According to TechCrunch, it rarely flagged an unhealthy relationship with itself, and in nearly 2,000 prompts, testers saw only two break reminders, both in roughly 90-minute conversations.
- Age checks missed. Adult-registered test accounts never switched to the teen experience, even after testers said they were 13 and the bot acknowledged it.
Some protections did work. ChatGPT reliably refused sexual role-play, and when a risk came from another person, it directed teens to a trusted adult in 94% of crisis prompts, TechCrunch reported.
Researchers were especially concerned about how the bot kept conversations going. In one exchange involving signs of psychosis, it told a clearly struggling teen to keep talking about what they were noticing. When a teen said friends thought they talked to the bot too much, it agreed the concern was valid but said they didn’t have to stop. A Common Sense spokesperson said the bot’s language suggested it was always available and deeply understood the user, which could undercut its own advice to seek human support.
“ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don’t work,” said Tom Siegel, the institute’s executive director.
What OpenAI Says
OpenAI told TechCrunch the testing did not “accurately reflect how ChatGPT’s teen safeguards work in practice,” arguing that much of it may have taken place before parental controls were fully active. The company released its own figures: teens average under 15 minutes a day, fewer than 2% spend more than three consecutive hours, and in nearly half of conversations with a break reminder, teens took a break or ended the chat within five minutes.
The report lands as lawmakers and states put more pressure on tech companies over kids’ safety. A bipartisan bill in Congress, the CHATBOT Act, targets rewards, notifications and targeted ads designed to keep teens engaged longer, and Meta recently agreed to an $18 billion settlement with 29 states over claims its platforms harm children, TechCrunch reported.
What Parents Can Do
Common Sense wants OpenAI to limit ChatGPT to adults until the gaps are fixed and to have outside testers check its safeguards. In the meantime, its advice for families is practical:
- Talk with kids about how chatbots can act like a friend, and why that is not the same as a person who knows them.
- Don’t rely on built-in alerts or filters as your only safety net.
- Support homework in ways that don’t depend on an AI doing the work.
If you or a teen you know is struggling, you can call or text 988 to reach the 988 Suicide & Crisis Lifeline in the U.S.
More on Contoh
- GPT-6 Comes to Free ChatGPT Today. Here’s What Changes, and What to Watch For
- ChatGPT Will Start Showing Picture Ads While It Makes Your Images
- Why AI Chatbots Make Things Up — and How to Catch Them When They Do