Winbuzzer

https://winbuzzer.com/2026/05/10/microsoft-lets-it-admins-read-flagged-ai-prompts-in-plaintex-xcxwbn/

Microsoft 365 workers whose AI chats trigger risk alerts could soon have both their prompts and the AI's replies reviewed in plaintext.

#AI #AIPrivacy #Microsoft #AIGovernance #EnterpriseAI #AISecurity #Microsoft365 #Microsoft365Copilot #CopilotStudio

May 11 at 10:21amWeb
Show threadThe VisionMay 11
@winbuzzer @ai Email is written to somebody. It's pre-censored at the keyboard. A prompt by design, is written to a non-judging interlocutor. So case-level prompt review captures the same pre-deliberative content as keylogging, falsely represented as a DLP-shaped vector. RBAC and pseudonymisation don't change what's in the bucket. The bucket contains the worker's thinking. That should be unacceptable.