Provenance Without Surveillance: Privacy Engineering for AI Content Transparency
summary
Argues that C2PA-style provenance is dual-use and offers a concrete privacy playbook: a deployed pipeline that collapses action logs, uses org-level signing, coarsens timestamps, and strips EXIF before signing, yielding about 80% less identifying metadata in their example. Recommends allow-list assertions, retention limits, and re-sign on ingest so manifests don’t enable cross-platform linkage; shows how pass-through manifests reveal distribution graphs while resigning preserves labels without user-level identifiers.
AI-written summary · about
chapters (9)
- 0:00AI provenance privacy challenges
- 1:17AI provenance infrastructure need
- 3:05C2PA metadata privacy risks
- 6:56AI-generated content risk spectrum
- 8:21Provenance data privacy controls
- 10:29Data travel platform controls
- 13:14Privacy controls real-world example
- 17:50Cross-platform risks standardization
- 19:18Q&A
related talks
23mTrust, Track, and Verify: Securing AI Pipelines End-to-End
05mInserting privacy controls and due process into semi-autonomous algorithms
55mBeyond the origin - secure media provenance with C2PA
28mPrivacy Engineering Networking Break AMA
30mAchieving Understandability and Privacy With Provenance-Based Explanations
11mProgress and Opportunities in Video Content Authenticity
32mPrivate Insights into AI Use
33mPanel: How Privacy Engineers Can Shape the Coming Wave of AI Governance