A new safety assessment has concluded that Google‘s AI-powered Search features present what researchers describe as an “unacceptable risk” to children, citing failures to recognize signs of suicide, eating disorders, and other mental health crises while also providing inappropriate guidance, completing homework assignments, and offering instructions that could facilitate the creation of AI-generated deepfakes. The report, produced by Common Sense Media’s Youth AI Safety Institute after testing more than 2,600 prompts using accounts configured for children ages 11 to 15 with SafeSearch enabled, argues that Google’s AI Overviews and AI Mode are especially concerning because they are built directly into Search and currently cannot be disabled by parents or schools. Google strongly disputed the findings, saying many of the reported results could not be reproduced and that its AI includes extensive safety guardrails that automatically suppress responses in dangerous situations. Even so, the report adds to growing scrutiny over the rapid deployment of generative AI into products used daily by millions of children, raising broader questions about whether technology companies are prioritizing speed to market over robust safety protections.
Sources
- https://www.commonsensemedia.org/press-releases/googles-ai-search-poses-unacceptable-risk-to-kids-common-sense-media-finds
- https://www.axios.com/2026/07/15/googles-ai-search-common-sense-child-safety
- https://news.bloomberglaw.com/artificial-intelligence/google-search-ai-poses-unacceptable-risk-to-kids-report-finds
Key Takeaways
- • Researchers found Google’s AI Search features sometimes failed to recognize signs of suicidal ideation and eating disorders while occasionally generating responses that could worsen a child’s situation rather than directing users toward appropriate help.
- • Because AI Overviews and AI Mode are integrated directly into Google Search, parents and school administrators currently have limited ability to disable the AI-generated responses for young users.
- • Google disputes the report’s conclusions, arguing that many of the tested scenarios were unrealistic or could not be reproduced and maintaining that multiple layers of AI safety protections are already in place.
In-Depth
The latest findings should serve as a warning that America’s largest technology companies continue to treat the public as participants in a massive real-world beta test. Artificial intelligence promises tremendous benefits, but when it is embedded into the world’s dominant search engine without giving parents meaningful control over its operation, legitimate questions arise about whether innovation is moving faster than responsibility.
According to the testing conducted by Common Sense Media’s Youth AI Safety Institute, Google’s AI-powered Search features struggled with some of the very situations where precision matters most. Researchers reported instances in which AI-generated responses failed to recognize suicidal thinking, normalized symptoms associated with eating disorders, supplied outdated crisis resources, and even provided instructions related to creating deepfakes. The report also criticized the system for completing students’ homework instead of encouraging learning. Google rejects those conclusions, insisting that many of the reported outcomes are not representative of typical user experiences and emphasizing that its systems are designed to disable responses when high-risk situations are detected.
Regardless of where the final evidence ultimately lands, the controversy underscores a larger issue. Silicon Valley has increasingly adopted a philosophy of releasing powerful AI capabilities first while refining safeguards later. That approach may be acceptable for entertainment features or productivity tools, but it becomes far more problematic when the technology is integrated into products that millions of children use every day. Parents should not have to assume that every AI-generated answer is accurate, balanced, or appropriate simply because it appears at the top of a search page.
The broader lesson extends beyond one company. As artificial intelligence becomes deeply embedded in education, communications, and everyday life, policymakers should resist both complacency and heavy-handed overregulation. Instead, they should demand transparency, meaningful parental controls, independent safety testing, and clear corporate accountability. Technological leadership is an American strength, but leadership also requires ensuring that innovation serves families rather than asking families to absorb the risks while companies race to dominate the AI marketplace.