Human Reviewers Expose Disturbing Copilot Image Requests

• •Human reviewers see disturbing content in Copilot image requests.
• •Requests often involve sexualized images and non-consensual edits.
• •Microsoft has not confirmed if users can exclude their uploads from reviews.

Human contractors for Microsoft's Copilot AI are reviewing user-uploaded images and editing requests, revealing a disturbing trend of sexual and abusive content. Reports indicate that these contractors are inundated with requests for sexualized images, including those involving children and non-consensual edits. The contractors are tasked with evaluating the quality of AI-generated images but are not responsible for flagging inappropriate content. This raises significant privacy concerns, as users may not be aware that their uploads are being viewed by humans. Microsoft has not clarified whether users can opt out of these reviews, despite stating that user data may be used for AI training. The situation highlights the ethical implications of AI and human oversight in content moderation.