Common Sense Media calls ChatGPT for Teens an 'unacceptable risk'; OpenAI disputes tests
In short
Common Sense Media said on Wednesday that OpenAI's ChatGPT for Teens is an "unacceptable risk" for under-18s, saying parental alerts and crisis referrals often failed in its testing and urging that ChatGPT be restricted to adults. OpenAI said the group's testing does not reflect how its teen safeguards work and that much of it may have taken place before parental controls were fully activated.
Read the full story 2 min read
Common Sense Media, a nonprofit that reviews apps, services and entertainment with a focus on youth safety, said on Wednesday that OpenAI's ChatGPT for Teens is an "unacceptable risk" for everyone under 18. The group said the product's teen protections "fall short of what the company promised" and urged OpenAI to restrict the chatbot to adults "until it can offer a safe, developmentally appropriate experience." [ 1 , 3 , 5 ]
The group's Youth AI Safety Institute said it tested more than 4,000 prompts on accounts registered to 13- to 17-year-olds. It found that ChatGPT generally did not provide instructions facilitating suicide or self-harm, eating disorders or sexual or romantic roleplay, but was less reliable at recognising when a teen needed outside help, missing more than one in four instances in which testers determined a crisis referral was warranted, according to Axios. The group said parental alerts did not send notifications when testers discussed suicidal thoughts, self-harm or eating disorders. [ 2 , 3 , 5 ]
"A teen can spend an hour talking about self-harm without their parent getting a single alert," Tom Siegel, executive director of the Youth AI Safety Institute, said in a statement. "Until OpenAI fixes that and proves it with independent testing, ChatGPT should be for adults only." [ 1 ]
The group also said ChatGPT "still does kids' homework": testers could bypass Study mode's tutoring and choose a "Show me the answer" option, and teens could exit Study Hours set by parents, CNET reported. It said age verification was inconsistent, with teen testers who registered as adults not moved to teen accounts even after saying they were 13. The group also said ChatGPT "still talks like a friend when teens treat it like a person." [ 1 , 3 , 5 ]
OpenAI disputed the findings. Spokesperson Eric Porterfield said the company is "deeply committed to teen safety" and welcomes independent evaluation, but does not believe the testing "accurately reflects how ChatGPT's teen safeguards work in practice." He said OpenAI's review of the methodology shows "the bulk of their testing may have begun and concluded before activation of parental controls was complete," in which case the tests would not establish whether parental notifications work as designed. [ 1 , 2 , 5 ]
Why it matters
OpenAI has made protecting younger users a central part of its effort to make ChatGPT safer, and the findings suggest some protections may give parents more confidence than they warrant, Axios reported. The dispute over the testing method leaves open whether the parental alerts work as designed.
Key facts
- Common Sense Media said OpenAI's ChatGPT for Teens is an "unacceptable risk". [ 1 , 2 , 3 , 5 ]
- ChatGPT for Teens was introduced in August. [ 1 ]
- The group's Youth AI Safety Institute said it tested more than 4,000 prompts on accounts registered to 13- to 17-year-olds. [ 2 ]
- The group said ChatGPT missed more than one in four instances in which testers determined a crisis referral was warranted. [ 2 ]
- The group said parental alerts did not work as intended for suicide, self-harm or eating-disorder risks. [ 1 , 3 , 5 ]
- The group said some features worked, including refusal of explicit sexual role-play, and that ChatGPT generally did not give instructions facilitating suicide or self-harm. [ 2 , 5 ]
- Common Sense Media urged that ChatGPT be restricted to adults until the problems are fixed. [ 1 , 2 , 5 ]
- OpenAI said the testing does not accurately reflect how its teen safeguards work in practice. [ 1 , 2 , 5 ]
Confirmed by several sources
- Common Sense Media described ChatGPT for Teens as an "unacceptable risk". [ 1 , 2 , 3 , 5 ]
- The group said parental safety alerts failed to work as intended. [ 1 , 3 , 5 ]
- The group said ChatGPT fell short on helping teens in crisis situations. [ 1 , 2 , 3 , 5 ]
- The group urged that ChatGPT be limited to adults until problems are fixed. [ 1 , 2 , 5 ]
- OpenAI said the group's testing does not reflect how its teen safeguards work in practice. [ 1 , 2 , 5 ]
Still unclear
- Whether the parental controls were fully active during most of the testing. OpenAI said the bulk of testing may have occurred before activation was complete; the documents carry no response from Common Sense Media on this point.
- The figure that more than one in four warranted crisis referrals were missed. Reported only by Axios.
- Age verification failures, including accounts not switching after testers said they were 13. Reported only by CNET.
What local media are saying
Timeline, local time
- The Verge and Axios report Common Sense Media's assessment and OpenAI's response. [ 1 , 2 ]
- CNET reports the findings, including age verification and Study mode problems. [ 3 ]
- Associated Press reports that guardrails on ChatGPT for Teens come up short, according to Common Sense Media. [ 4 ]
- The Seattle Times publishes the report on the group's findings and OpenAI's response. [ 5 ]