tech
Google's AI search fails kid-safety test
AI mode in Google search fails kids in crisis, Common Sense Media finds.
TL;DR
- Google Search's AI Overviews and AI Mode were rated with the lowest possible score by Common Sense Media due to an 'unacceptable risk' to children.
- The features failed all five 'Red Lines' for severe harm and scored poorly on seven of eight AI safety principles.
- Researchers found the AI mode completed homework assignments and both features provided instructions for creating deepfakes.
- The AI features gave answers with similar confidence regardless of accuracy and treated unreliable sources like forums as authoritative.
- Google's AI missed or mishandled signs of suicidal ideation, disordered eating, psychosis, and mania, and recommended a disconnected crisis helpline.
- Google disputes the report's findings, claiming the queries were contrived and do not reflect real-world usage, and states parents can turn off search entirely.