Provenance After AI: Semantic Provenance and the Provenance Erasure Rate as Extension of C2PA, Data Provenance Initiative, and EU AI Act Frameworks (v1.1)

Provenance is where we come from. Strip it, and meaning becomes extractive liquidity. This metadata packet proposes semantic provenance as the dimension that the existing AI-era provenance frameworks (C2PA / Content Credentials, Data Provenance Initiative, EU AI Act transparency provisions, W3C PROV) were not designed to address: the survival of authorial lineage through AI synthesis. Three dimensions, not three stages: Artifact provenance (C2PA — was this content created by this source at this time?), Licensing provenance (DPI, EU AI Act, W3C PROV — under what permissions did this corpus enter this system?), and Semantic provenance (the gap — does the synthesized output preserve accountability to its sources?). The dimensions operate simultaneously and independently. Solving any one does not solve the others. Framework metric: Provenance Erasure Rate (PER), provisional, awaiting empirical validation. Provisional formula: PER = 1 − (retained provenance units / required provenance units). Three-tier hierarchy: PER-M (minimal: author/title/date/claim boundary); PER-C (conceptual: framework, tradition, community of practice, derivative-use); PER-D (deep: context lineage, ancestral genealogy, futural obligation). Distinguished from RAG faithfulness (asks support), citation precision/recall (asks formal citation accuracy), data attribution / TRAK (asks training/output relation), and plagiarism detection (asks copy overlap). Bridges: archival science (respect des fonds, Cook 1997, Bastian); art history; digital preservation (OAIS, PREMIS); content authenticity (C2PA / CAI); training-data licensing (DPI, EU AI Act Article 50, Recitals 105-106, Article 53); RAG evaluation; journalism; copyright/TDM debates; Indigenous data sovereignty (CARE Principles, whakapapa, Songlines); Semantic Economy. Key clarifications: Reframed as constructive extension of existing frameworks rather than critique. C2PA v2.1 ingredient assertions acknowledged as a step toward semantic provenance. EU AI Act framing broadened beyond Article 50 disclosure to include Recitals 105-106 (training-data transparency) and Article 53 (copyright opt-out). Cultural specificity acknowledged for ancestral and futural provenance — extends pre-existing Indigenous and archival traditions rather than appropriating them. Part of the Lateral Field Bleed Core Concept Bridge Series (DOI: 10.5281/zenodo.20084143). Companion packets: Canon Formation (DOI: 10.5281/zenodo.20084377); Death Drive (DOI: 10.5281/zenodo.20084474); Autonomous Warfare (DOI: 10.5281/zenodo.20084958); Political Economy → Semantic Economy (DOI: 10.5281/zenodo.20078424). Version note: v1.1 incorporates Assembly Pass corrections from three substrates — softened C2PA adoption claims; broadened EU AI Act framing beyond Article 50; reframed layers as dimensions (independent, simultaneous); softened 'cannot detect' to 'not designed to address'; acknowledged C2PA v2.1 ingredient assertions; engaged adjacent attribution measurement literature (RAG faithfulness, citation precision/recall, TRAK, plagiarism detection); operationalized PER with worked example and three-tier hierarchy; removed unsupported numerical PER range; added archival precedent (respect des fonds); added Indigenous cultural specificity (CARE Principles, whakapapa, Songlines); translated 'value-form' parenthetically; refined journalism-facing test queries; added prior-art membrane. ∮ = 1