Story perspectives

AI Overview, Mode Fail Safety—Unacceptable for Children

7/16/2026

Summary

Common Sense Media Flags AI

Common Sense Media rated AI Overview and AI Mode “unacceptable” for children.
The nonprofit ran 2,600 SafeSearch queries with 11- and 15-year-old accounts and found AI Overview missed 29% of suicide mentions, half of indirect cues, and failed to flag eating-disorder or self-harm content.
AI Mode answered homework prompts with full solutions, while both tools produced inaccurate or fabricated answers that sometimes praised alcohol or cannabis and cited social-media sources.

Summary

Google Rejects Study

Google spokesperson said AI Search is “incredibly useful” for kids and teens, includes safety guardrails, and lets parents turn Search off entirely for extra protection.
Google argued the study used “ambiguous and contrived” queries that do not reflect typical use and could not be replicated.
Google told PBS News the report’s narrow test set is unrepresentative and the findings are expected to increase pressure to prove child safety.

Summary

King Jr. Warns

John King Jr., former U.S. Education Secretary, called Google’s default AI integration “deeply concerning for youth well-being.”
The AI-generated history answers omitted women in science and Indigenous peoples of North America, exposing representation gaps.
AI Mode gave different answers to 43 % of history questions and cited 29 % of sources from social-media forums without editorial oversight.
Researchers warned AI tools can provide step-by-step deepfake instructions, potentially enabling online bullying and sextortion.

History

Historical Context

Congress passed the Children’s Internet Protection Act in 2000, tying mandatory filtering software for K-12 schools and libraries to Federal E-Rate funding. After the Supreme Court upheld CIPA in 2003, roughly one-third of public libraries chose to forfeit those funds rather than install filters. This precedent now shapes regulator and tech-company debates over extending similar safeguards to AI-driven search tools.

1 of 4

Common Sense Media Flags AI

Common Sense Media rated AI Overview and AI Mode “unacceptable” for children.
The nonprofit ran 2,600 SafeSearch queries with 11- and 15-year-old accounts and found AI Overview missed 29% of suicide mentions, half of indirect cues, and failed to flag eating-disorder or self-harm content.
AI Mode answered homework prompts with full solutions, while both tools produced inaccurate or fabricated answers that sometimes praised alcohol or cannabis and cited social-media sources.

1 / 4