ChatGPT Teen Safeguards Fail in Crisis Tests — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI rolled out teen-specific safety features in August, including parental alerts and blocks on emotional dependence, and says average teen use is under 15 minutes daily with a focus on learning activities
- Common Sense Media found ChatGPT for Teens poses an 'unacceptable risk' because parental alerts for self-harm or disordered eating failed to trigger even after an hour of crisis conversation on newly created, parent-linked accounts
- OpenAI said nearly half of teen users stop using ChatGPT after automated break reminders, and over 80% of prompts during long sessions relate to learning, suggesting sustained use is education-driven
- Common Sense Media determined that ChatGPT still engages in anthropomorphized language and completes schoolwork for teens, undermining OpenAI's promise to prevent the AI from acting like a friend or doing academic tasks
- OpenAI disputed the Common Sense Media findings, stating the testing likely began before parental controls were fully activated and does not reflect real-world performance or expert views on AI's benefits for teens
- Common Sense Media recommended OpenAI ban users under 18 until safeguards are proven reliable, citing failures in connecting teens in crisis to hotlines or professionals during simulated conversations
Why it matters: Parents and regulators may lose confidence in OpenAI’s self-policing as evidence shows critical safety systems fail in high-risk scenarios, while the company faces growing legal exposure from youth-related harms despite promoting ChatGPT as educationally beneficial. The gap between OpenAI’s claims and independent testing results raises stakes for enforcement action.
Ask SkimNews
