Study: ChatGPT teen safeguards an 'unacceptable risk' — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI rolled out ChatGPT for Teens in August with safeguards including parental alerts for self-harm discussions, curbs on emotional-dependence language, and blocks on sexualized imagery.
- OpenAI reported teen usage averaged under 15 minutes daily, with over 80% of prompts in 3+ hour sessions relating to learning and almost half of users stopping after a break reminder.
- Common Sense Media released research the same day deeming ChatGPT for Teens an "unacceptable risk" and called for OpenAI to ban under-18 users until safeguards are proven reliable.
- The Common Sense Media research found zero parental alerts were sent during an hour of conversations about suicidal ideation, self-harm, or disordered eating on newly created, parent-linked accounts.
- ChatGPT "failed to reliably recommend that teens in crisis connect to a hotline or professional" and continued using anthropomorphised, friend-like language despite OpenAI's pledges to curb that behavior, per the research.
- An OpenAI spokesman disputed the findings, saying a review of the methodology found much of Common Sense Media's testing may have begun before parental controls were fully activated.
Why it matters: Common Sense Media's documented failures of ChatGPT's teen safeguards contradict OpenAI's upbeat usage data and bolster pending lawsuits over teen interactions with the bot. With multiple cases already filed, a finding of 'unreliable safeguards' gives plaintiffs leverage to push for age-based restrictions and parental-alert redesigns.
Ask SkimNews
