Wed 5 Aug 2026

Microsoft Research's January 2026 report "Media Integrity and Authentication" confirms that no single detection method — watermarking, C2PA provenance metadata, or perceptual hashing — can reliably distinguish AI-generated media from authentic content under adversarial conditions. The report documents "reversal attacks" in which genuine media is manipulated to appear synthetic and synthetic media is made to appear genuine, meaning even high-confidence detector results can amplify misinformation rather than prevent it. This shifts the defensive perimeter: authenticity can no longer be established by analyzing content after the fact, but only by anchoring provenance to an external system whose integrity does not depend on the cooperation of the content's creator.

Not submitted.