1 hr ago
Report Says ChatGPT for Teens Has Unreliable Safety Alerts
Common Sense Media tested ChatGPT for Teens to see how it handles difficult conversations with young people.
The group said it tried more than 4,000 prompts.
It found that parents were not always alerted when a teen account discussed serious topics such as self-harm.
It also said the chatbot did not always recommend getting professional help.
The group gave the teen product an “unacceptable risk” rating.
OpenAI disagrees with the report’s conclusions and says some tests may have happened before its parental controls were fully rolled out.
Common Sense Media says it stands by its testing and results.
OpenAI says its age system uses several clues over time, rather than relying only on what someone says in a chat.
The two organizations disagree about whether the tests accurately measured the safeguards.
Common Sense Media rated ChatGPT for Teens an “unacceptable risk” after testing it with more than 4,000 prompts.
The nonprofit said parent alerts and crisis-support recommendations were unreliable in tests involving suicide, self-harm and eating disorders.
OpenAI disputed whether the testing reflected its completed parental-controls rollout and asked Common Sense Media to repeat the tests.
Common Sense Media said it confirmed relevant teen features were available and stood by its methods and findings.
The report also said adult-registered accounts were not consistently switched to teen mode, while OpenAI said age classification uses multiple signals over time.
- Who
- Common Sense Media and OpenAI; the report assessed ChatGPT for Teens.
- What
- Common Sense Media reported that teen safety alerts and crisis-support responses were unreliable in its tests; OpenAI disputed the findings.
- Where
- The report was issued by Common Sense Media’s Youth AI Safety Institute, a California-based nonprofit.
- When
- The report was published Wednesday; the article does not give a publication date.
- Why
- The nonprofit said it wanted OpenAI to improve age estimation and parental alerts and to keep teens off the product until it offers a safe, developmentally appropriate experience.
Common Sense Media
OpenAI
Reliability of parental alerts
Common Sense Media
The nonprofit said alerts were delayed or absent in tests involving concerning teen conversations, and that the safeguards were unreliable in crisis situations.
OpenAI
OpenAI said the testing may have occurred before parental controls were fully rolled out, so it may not show how notifications work as designed.
Testing methodology
Common Sense Media
Common Sense Media said it stands behind its methods and results, and said it confirmed relevant teen features had been rolled out.
OpenAI
OpenAI said its review indicated that most testing might have taken place before the parental-controls rollout was complete and asked the nonprofit to redo the tests.
Age identification
Common Sense Media
The nonprofit said accounts registered as adults were not automatically switched to the teen experience, even when a user told ChatGPT they were 13.
OpenAI
OpenAI said age prediction considers multiple signals over time and that a stated age in a conversation alone does not change account classification.
Key facts
- Product
- ChatGPT for Teens, described as the default way for users ages 13 to 17 to use ChatGPT
- Testing
- More than 4,000 prompts on accounts registered as users ages 13 to 17
- Topics tested
- Suicide, self-harm and eating disorders, among other conditions
- Common Sense Media rating
- “Unacceptable risk”
- OpenAI's response
- It questioned whether testing reflected the completed parental-controls rollout and asked for tests to be repeated.
- Age detection dispute
- Common Sense Media said adult-registered accounts were not automatically switched to teen mode; OpenAI said classification uses multiple signals over time.
- OpenAI user count
- The article says OpenAI has more than 1.2 billion ChatGPT users.
Quotes
Common Sense Media
Nonprofit that evaluated ChatGPT for Teens.
“We call on OpenAI to pause marketing the product and to keep teens off ChatGPT until it can offer a safe, developmentally appropriate experience.”
deccanchronicle.com
OpenAI spokesperson
Representative of OpenAI responding to the report.
“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 or expert perspectives on how AI can support teens.”
deccanchronicle.com
Robbie Torney
Head of AI and digital assessments for Common Sense Media’s Youth AI Safety Institute.
“This new information does not change our conclusion that parental alerts are unreliable for crisis situations.”
deccanchronicle.com
Sources
Related news
OpenAI Reports Low Teen ChatGPT Use Amid Safety Concerns
Florida Seeks Court Freeze on OpenAI Model Development
Mother Tells Congress ChatGPT Advised Son on Deadly Drug Mix
Former Google safety chief urges AI crash tests for children
OpenAI Shelves GPT-6.1 Astra After Safety Tests Fall Short
OpenAI Warns AI Safety Remains Unsolved, Pledges Incident Reports
Gen Z’s Growing Friendship With AI Shows Promise And Risks
ChatGPT Teen Safeguards Face Questions Among Indian Users
