A new analysis by Common Sense Media concludes that Google’s AI-enabled search features expose children to unacceptable risks, citing failures to detect self-harm and substance use, to flag psychosis-like statements, and to prevent academic cheating. Testing more than 2,600 interactions using child accounts, the group found Google’s AI Overview and AI Mode frequently gave inaccurate, unsafe or inconsistent answers, and completed all assigned homework prompts. The tools, which are embedded in Search and can’t be disabled, sometimes celebrated underage drinking and offered guidance on deepfakes and voice cloning, the report said. Google defended its products as useful and protected by multiple safety layers, arguing the report used contrived queries that don’t reflect typical behavior, and noted parents can manage controls or turn off Search entirely. The findings arrive as U.S. lawmakers weigh new rules for children’s AI use and states tighten student data privacy. Advocates urge Google to let schools and parents switch off AI Overviews; some educators say they’re steering students away from standard Google searches until stronger safeguards are in place.
Related articles:
– Online Safety Act 2023 (UK)
– Google Safety Center: Safety features and parental controls
