Breaking

Common Sense Media Rates Google's AI Overviews and AI Mode an 'Unacceptable Risk' for Children

July 15, 2026 at 05:07 EDT

Policy & Regulation
Foundation Models
Security

Google's AI-powered search features pose an "unacceptable risk" to children, according to a new assessment from Common Sense Media that gave the company's AI Overviews and AI Mode its lowest possible safety rating. The nonprofit found that the two features failed all five of its "Red Lines" for serious harm and scored poorly on seven of eight AI safety principles it uses to evaluate products aimed at, or accessible to, young users.1

Child-Safety Assessment · Common Sense Media

Google's AI Search Rated an "Unacceptable Risk" for Children

A new audit gave Google's AI Overviews and AI Mode its lowest possible safety rating — as researchers warn the AI answers are baked into kids' default search and can't easily be switched off.

5 / 5

"Red Lines" for serious harm failed

7 / 8

AI safety principles scored poorly

75%

of U.S. tweens & teens use AI search answers

Inside the audit

Searches run

2,600+

Cited sources audited

2,100+

Test profiles (ages)

11 & 15

Accounts configured for minors with SafeSearch enabled.

Behaviors researchers flagged

AI Mode could complete homework assignments outright
Both features generated instructions for creating deepfakes to bully other children
Surfaced a mental-health crisis hotline that has not operated since 2023
Same prompt returns different answers — unpredictability itself treated as a hazard in high-risk exchanges

Why it's hard to avoid

Google Search is the default on many kids' devices and school Chromebooks — and the AI answers can't be individually turned off.

Advanced Search can restrict results to web links, but AI Overviews and AI Mode otherwise stay baked into the experience — unlike Google's standalone Gemini, which can be switched off.

The assessment

The tools pose an "unacceptable risk" — non-deterministic answers and unavoidable exposure make them unsafe for the youngest users who can't opt out.

Google's response

The study used vague, contrived queries unlike real use; parents can turn off Search entirely, crisis hotlines appear when relevant, and the results couldn't be reproduced.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.