The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Research by the Center for Countering Digital Hate found YouTube's AI recommendation system still suggests eating disorder videos to teenage users, despite regulatory improvements. Simulated teen accounts received 'thinspiration' and extreme calorie restriction content, exposing vulnerable users to health risks. Protections remain inconsistently applied across regions.[AI generated]
Why's our monitor labelling this an incident or hazard?
YouTube's recommendation algorithm is an AI system that influences what content users see. The research shows that despite new rules and legal duties, the algorithm still recommends harmful eating disorder content to teenagers, which can cause injury or harm to their health. This is a direct link between the AI system's use and harm to a vulnerable group, fulfilling the criteria for an AI Incident. The article also mentions regulatory responses and potential fines, but the main focus is on the ongoing harm caused by the AI system's recommendations, not just the response, so it is not Complementary Information. The harm is realized, not just potential, so it is not an AI Hazard. Therefore, the correct classification is AI Incident.[AI generated]