Human contractors reviewing Microsoft's Copilot AI are exposed to disturbing user requests and generated images, including explicit content and images related to eating disorders. These contractors are tasked with evaluating the AI's success in fulfilling prompts, such as generating images with specific sexual characteristics. The content they encounter is described as lewd and sexually explicit, raising concerns about the well-being of these reviewers. AI
IMPACT Highlights the ethical and psychological toll on human reviewers moderating AI-generated content, potentially influencing future AI safety protocols.
RANK_REASON News about the operational issues and content moderation challenges of an existing AI product.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →