Skip to content
Longterm Wiki
Back

Microsoft Video Authenticator

web
Microsoft·blogs.microsoft.com/on-the-issues/2020/09/01/disinformati...

Credibility Rating

4/5
High(4)

High quality. Established institution or organization with editorial oversight and accountability.

Rating inherited from publication venue: Microsoft

Relevant to AI safety discussions around misuse of generative AI; illustrates industry-led technical countermeasures to deepfake disinformation, with implications for AI deployment norms and content provenance standards.

Metadata

Importance: 42/100press releasenews

Summary

Microsoft introduces Video Authenticator, a tool that analyzes images and videos to detect AI-generated manipulations (deepfakes) by identifying subtle blending boundaries and grayscale elements invisible to the human eye. The initiative is part of a broader effort including partnerships with NewsGuard and media literacy campaigns to combat disinformation ahead of the 2020 U.S. election. Microsoft also introduced a content provenance system to help publishers and journalists signal content authenticity.

Key Points

  • •Video Authenticator provides real-time confidence scores on whether media has been artificially manipulated, detecting subtle deepfake artifacts.
  • •Microsoft partnered with NewsGuard and other organizations to distribute the tool to news organizations and campaigns to counter election disinformation.
  • •A content provenance technology was introduced to cryptographically certify the origin and history of media content.
  • •The tool is part of Microsoft's AI for Health and broader responsible AI deployment, acknowledging deepfakes will improve and detection must evolve.
  • •Microsoft emphasized media literacy alongside technical tools, recognizing that technology alone cannot solve the disinformation problem.

Review

Microsoft's approach to addressing disinformation represents a multi-faceted strategy combining technological innovation and educational initiatives. The Video Authenticator, developed by Microsoft Research and the Responsible AI team, provides a real-time confidence score for detecting artificially manipulated media by analyzing subtle visual cues that might escape human perception. The technology acknowledges its own limitations, recognizing that AI detection methods are not infallible and will need continuous evolution. Microsoft's comprehensive strategy extends beyond technical solutions, including partnerships with media organizations, academic institutions, and initiatives like Project Origin and media literacy programs. By collaborating with entities like the AI Foundation, BBC, and University of Washington, Microsoft aims to create a holistic approach to combating synthetic media and disinformation, emphasizing both technological detection and public education.

Cited by 1 page

PageTypeQuality
AI-Accelerated Reality FragmentationRisk28.0
Resource ID: 97907cd3e6b9f226 | Stable ID: ZmRiYjJhNT