San FranciscoTeen ChatGPT skipped crisis referrals in a quarter of tests.
Common Sense Media, a US youth-safety non-profit, rates the product an unacceptable risk to children.
Its 4,000-plus prompts covered suicide, self-harm, disordered eating and impaired reality, and linked parent accounts received zero alerts.
OpenAI says most testing may have ended before parental controls were active, so the tests would not show whether alerts work.
Common Sense Media says it confirmed the features were live before testing, and stands by its results.
How each outlet framed it
drawn from 90+ reports worldwide
- PC Mag Middle East leans critical
- flags lack of Middle East region-specific rollout info despite Common Sense Media's documented safety failures
Sources: PC Mag Middle East