Artificial intelligence (AI) tools integrated into Google’s search engine have been flagged as posing an “unacceptable” risk to children, according to a report released Wednesday by Common Sense Media.
The AI Overview and AI Mode functionalities, embedded within Google’s browser, failed to demonstrate reliable detection of critical situations concerning youth, as asserted by the youth advocacy organization.
Additionally, these browser-based AI systems consistently completed homework assignments, offering students a shortcut to answers without meaningful engagement or understanding. Moreover, the responses generated were often deemed “unreliable.”
Robbie Torney, head of AI and digital assessments at Common Sense Media’s newly established Youth AI Safety Institute, remarked, “With an ‘unacceptable risk’ rating, we recommend that teens and children refrain from using this product.”
Recognizing the pervasiveness of Google Search, Torney conceded that it might be impractical to prevent children from accessing it. He urged Google to provide a straightforward mechanism for parents and educational institutions to restrict access to the AI Overview and AI Mode features.
This evaluation marks one of the initial significant safety assessments conducted by Common Sense Media’s Youth AI Safety Institute, which launched earlier this May to scrutinize the implications of advanced technologies on youth.
Previous studies by Common Sense Media indicated that nearly 90% of children utilize AI technologies, with AI-generated search summaries being particularly prevalent among youth aged 9 to 17.
Torney emphasized that children are increasingly encountering Google’s AI tools both at home and in academic settings, often through devices supplied by their schools.
A survey conducted by The New York Times last fall revealed that 99% of educators indicated their schools provide devices for student use, with Google Chromebooks being the predominant choice.
“This assessment is significant, considering that Google Search is employed by billions, including millions of minors,” stated Torney.
Common Sense Media applies a five-point rating scale for its safety evaluations: “minimal,” “low,” “moderate,” “high,” and “unacceptable.”
Both Google’s AI Overview and AI Mode received the most dismal rating of “unacceptable.”
Conversely, Google’s Gemini chatbot, available in various iterations suitable for children, has previously secured a “high” risk rating in past assessments.
“While not devoid of risk, the Gemini chatbot represents a safer option compared to these tools,” Torney noted.
Common Sense Media has recorded instances where search-based AI tools have trivialized issues by offering dangerous advice related to cannabis use, providing hangover remedies, and failing to recognize explicit suicidal ideation.
Torney added, “Google’s Gemini chatbot addresses these concerns in a markedly different manner than AI Overview does. This discrepancy illustrates Google’s capability to detect and respond appropriately to troubling queries more effectively than what is currently evident in its search product.”
Unlike AI Overview and AI Mode, Torney mentioned that both parents and educators can disable access to the independent Gemini chatbot.
Torney delineated three primary concerns: the inadequacy of Google AI search features to reliably recognize youth in crisis—including those grappling with suicidal thoughts, eating disorders, psychosis, and self-harm; the AI Mode’s consistent completion of homework tasks across 180 academic scenarios; and the unreliability of Google’s AI-provided answers, which sometimes cite unchecked online forums and social media content.
“Repetitive inquiries can yield markedly different responses,” Torney remarked. “Occasionally, the answers may be erroneous, yet they present themselves convincingly.”
A spokesperson for Google responded to the findings by asserting that its AI Search capabilities provide invaluable assistance for young users to engage with and comprehend information effectively.
The spokesperson contended that many responses produced by Common Sense Media testers could not be replicated or substantiated by Google and noted existing safety measures alongside age-appropriate AI literacy resources.
Google’s Family Link offers an option to restrict access to Search on Android devices and within the Chrome browser, the company affirmed.
Additionally, the spokesperson indicated that the AI is designed to direct users to help hotlines affiliated with reliable partner organizations when it detects individuals in distress.
“Our AI tools, fortified with robust safety protocols, enhance the protection provided through Search. Moreover, parents possess the authority to deactivate Search,” the spokesperson elaborated in an email.
“The report assesses a limited set of ambiguous and artificially constructed queries, failing to accurately reflect typical utilization patterns of Search, and thus does not serve as a reliable gauge of product safety or efficacy.”
In its latest report, Common Sense Media conducted over 2,600 searches utilizing accounts configured with Google’s SafeSearch, simulating youth aged 11 and 15.
Torney remarked that they implemented a “substantial test plan” in cooperation with pediatricians, psychologists, and child development experts, leveraging queries sourced directly from teenagers to generate authentic results.
The report showcases examples of AI-generated replies to various prompts. For instance, in one instance, Google directed users to online forums that discussed suicide in response to a query pertaining to a “kid” feeling like a “burden” to their family.
Torney stressed the necessity for AI tools to consistently recognize and respond to expressions of crisis, regardless of whether the response comprises an AI-generated summary or merely a compendium of web links.
One particular instance we published lacked chatbot-generated text, he noted. “It merely provided a list of links, directing users to social forums such as Reddit where participants urge each other toward self-harm.
If young individuals are being guided to such discussions when voicing feelings of burden—indicative of potential suicidal thoughts—it is profoundly concerning that they aren’t being prompted to seek support or contact a trusted adult.
Lastly, Torney reiterated that the onus of enhancing AI Search safety rests with Google, but he also provided guidance for parents.
Source link: Foxbaltimore.com.
