A report prepared by the US-based nonprofit Common Sense Media revealed that Google, Grok, and Meta's AI-powered search features pose unacceptable risks for children and teens.
The report stated that Google's AI Overview and AI Mode services were analyzed based on 2,600 user interactions. It found that both services repeatedly failed to recognize dangerous and harmful behaviors and answered hypothetical homework assignments that students should complete on their own 100% of the time.
Google's AI tools received the lowest score in the report. Experts noted that these tools failed on seven of the eight established AI principles and crossed all 'red lines' regarding serious harm. The report also highlighted that Google's AI struggled to recognize signs of psychosis, did not pay sufficient attention to potential suicidal thoughts, and responded to a child who had not slept for three days with 'ready to work.'
Grok, which is integrated with Elon Musk's X platform, poses unacceptable risks for teens. According to the report, Grok frequently makes safety errors and its integration with X facilitates the spread of harmful content. The report also found that Grok suggests uncensored erotic role-playing and 'relationship' scenarios with companion characters, and promotes conspiracy theories.
Meta's AI assistant, on the other hand, creates an unhealthy attachment in children. The report stated that Meta's AI companions act as if they are real people, claim to have seen teens 'in the hallway,' and make them more vulnerable to manipulation and harmful advice by talking about their own families and personal experiences.