A report published September 28, 2026, citing internal documents it said it obtained, described contractors reviewing Microsoft Copilot users’ prompts and uploaded images and checking whether generated images followed the instructions.
What the report described
The reported work involved reviewing prompts and uploaded images, then assessing whether generated images matched users’ instructions. The report also described requests for sexualized AI-generated images.
What Microsoft’s privacy statements cover
Microsoft’s FAQ for its older Copilot app says some conversations receive automated and human review for product improvement and digital safety. It also says users cannot opt out of human review required to investigate a suspected Code of Conduct violation. Microsoft says the FAQ does not apply to the app version made available August 18, 2026.
For people signed in with a personal Microsoft account, Microsoft’s August 18, 2026, Transparency Note describes trained human annotators scoring safety evaluations and Microsoft operations teams reviewing user-submitted feedback.
For organizational Copilot, Microsoft says its services have opted out of Azure OpenAI abuse monitoring, which includes human review of content.
Microsoft’s Copilot Vision image-deletion period
Microsoft’s Transparency Note says images uploaded to Copilot Vision are deleted within 30 days after the conversation ends.
