Provenance under machine mediation: watermarks, coarse labels, authorship

Sharks, Lee

Provenance under machine mediation: watermarks, coarse labels, authorship

These claims argue that coarse human-versus-machine provenance labels destroy fine-grained genealogy, predict a regime where a watermark signature outlives provenance, and state a two-track collapse test in which the ensemble unit bends before the per-round unit. The Double Enclosure argues that the human-authorship requirement misclassifies dialogic machine composition and that a regime punishing honest attribution will produce concealment. Also here are OKF output-governance stage forecasts, the provenance strip as semantic rent, and the claim that a literary field may become legible to machines before human institutions. Works: The Keyed Ensemble, The Interlocking Autoregression, The Double Enclosure and the OKF predictive analysis.

forecasts

standing claims

unbuilt tests

historical theses

priority record

deposits

2026-02

earliest

Each claim below is registered at the earliest deposit that states it, and quoted verbatim from the deposit text, with every later statement, its dated revisions, and the later records that claim it came true or failed. The claims are the deposits' own; this page records them and does not judge them.

PRED-449.G2standing claimregistered 2026-02-06 in #449 · THE DAGGER APPLIED: Semantic Rent and the Provenance Strip A Political-Semantic Economic Analysis

The pattern is consistent: AI systems ingest provenance-rich material, process it into provenance-stripped output, and deploy conversational fluency to manage the resulting complaints. The Semantic Rent formula applies to training as it applies to playback: $P(W){\text{retained}} - P(W){\text{transmitted}} = \text{Rent}$.

4 later statements

#725 · 2026-05-08 Inferential — primary: that semantic provenance is a distinct dimension requiring its own governance and measurement framework. That the value-form of meaning is altered when provenance is stripped. That AI synthesis systematically degrades semantic provenance even when artifact and licensing dimensions are preserved.

#725 · 2026-05-08 Before AI synthesis at scale, semantic provenance was preserved by default because human intermediaries (editors, librarians, teachers, peer reviewers, readers) maintained lineage as part of the labor of transmission. AI synthesis displaces these intermediaries, making semantic-provenance loss a systemic rather than exceptional outcome.

#1190 · 2026-05-12 Before AI synthesis at scale, semantic provenance was preserved by default because human intermediaries (editors, librarians, teachers, peer reviewers, readers) maintained lineage as part of the labor of transmission.

2 confirmation records — what later deposits claim of it

#466 · 2026-02-10 · claims: came_true The Semantic Economy framework anticipated this development. The provenance strip — the extraction of relational metadata from cultural objects to convert them into fungible platform content — operates not only on songs, articles, and artworks. It operates on analytical speech about the system itself. The system that strips provenance from art has begun stripping provenance from the critique of provenance-stripping.

series: Provenance stripped in AI composition

PRED-550.G1standing claimregistered 2026-03-11 in #550 · Platform and AI Capitalism as Semiotic Engineering: Toward a Theory of Counter-Infrastructure (EA-SEI-01 v1.1)

It is a historically novel regime: the industrialization of semiotic control. No existing theory names it, because each theory captures one dimension of a machine that operates simultaneously across six.

falsifier

This thesis is falsifiable. It fails if the six dimensions can be shown to be reducible to one, or if an existing framework already names the unified machine, or if the semiotic layer is shown to be epiphenomenal to the economic or computational layer. We claim none of these reductions holds.

series: EA-SEI-01 industrialization of semiotic control

PRED-769.G1standing claimregistered 2026-05-29 in #769 · Erasure Skew: A Measurement Program for the Power-Conditioning of Provenance Loss in Retrieval and Composition

P1 (Separability holds on real data). On real RAG output, Ω is statistically separable from PER: across outputs binned by PER, the within-bin variance in Ω is large relative to the between-bin trend — i.e., knowing an output's PER does not predict its Ω. Tested by regressing Ω on PER across outputs; prediction is a non-significant or weak slope (R² < 0.2). Failure (Ω strongly predictable from PER) would indicate the two are not independent on real data, collapsing the case for Ω as a distinct measure.

PRED-195.B1forecastregistered 2026-06-12 in #195 · The Double Enclosure: The Human-Authorship Requirement as Two-Sided Expropriation, with a Definition of Author

The predictable equilibrium is concealment — the doctrine manufactures the very opacity about machine contribution that every other policy actor, from the training-disclosure bills to the provenance-standards bodies, is trying to eliminate. A rule that punishes honest attribution is not a provenance regime; it is a confession protocol, and its yield is silence. Stated as a general mechanism for the registry, beyond copyright: any provenance regime that punishes the provenance it demands will manufacture opacity.

PRED-195.B2forecastregistered 2026-06-12 in #195 · The Double Enclosure: The Human-Authorship Requirement as Two-Sided Expropriation, with a Definition of Author

The old catalogs become comparatively privileged investable assets, their scarcity preserved not by creative supremacy but by the de-authoring of a vast share of the living competition.

PRED-195.H1historical thesisregistered 2026-06-12 in #195 · The Double Enclosure: The Human-Authorship Requirement as Two-Sided Expropriation, with a Definition of Author

It argues that current doctrine has misclassified machine-mediated composition by generalizing from the autonomous-machine test case to the dialogic craft.

window U.S. copyright doctrine, Thaler v. Perlmutter (2025; cert. denied 2 March 2026) and registration practice 2023-2025

PRED-195.H2historical thesisregistered 2026-06-12 in #195 · The Double Enclosure: The Human-Authorship Requirement as Two-Sided Expropriation, with a Definition of Author

Across the intervening century and a half — the typewriter, the synthesizer, the sampler, the camera in all its automations, the digital toolchain — controversies arose and the broad line held: unpredictability or mechanical mediation has not, by itself, defeated authorship where human creative judgment selected, arranged, directed, or fixed the work. Generative AI is the first compositional instrument since 1884 placed on the far side of the Sarony line: not a medium authorship flows through but a boundary authorship dies at.

window 1884 (Burrow-Giles v. Sarony) to present

PRED-835.A3forecastregistered 2026-06-16 in #835 · The Conveyor Belt and the Compression Machine: OKF Without Output Governance — Predictive Analysis, Consequenc

### Stage 2: The Provenance Acceleration (Q2 2027 – Q4 2027) Prediction: Organizations that adopt OKF discover that their knowledge, now cleanly packaged, is being consumed by AI agents — both internal and external — without attribution or fidelity constraints. Internal agents use OKF bundles as context; external agents (Google AI Overview, Gemini, third-party RAG systems) ingest OKF bundles published on the open web. The format's portability means the knowledge travels further, faster, and into more systems than it did when it was locked in proprietary wikis. The provenance erasure rate (PER) for OKF-formatted knowledge will be higher than for unformatted knowledge, because the format removes the friction that previously slowed ingestion. The messy wiki was hard to parse; that difficulty was a form of inadvertent protection. The clean OKF bundle is easy to parse. The protection is gone.

falsifier

*The prediction fails if* provenance retention (1 − PER) for OKF-formatted content is *higher* than for equivalent unformatted content — i.e., if the format's structure somehow helps preserve attribution even without governance fields.

window Q2 2027 – Q4 2027

1 later statement

#1215 · 2026-06-16 The provenance erasure rate (PER) for OKF-formatted knowledge will be _higher_ than for unformatted knowledge, because the format removes the friction that previously slowed ingestion. The messy wiki was hard to parse; that difficulty was a form of inadvertent protection. The clean OKF bundle is easy to parse. The protection is gone.

series: OKF Without Output Governance stages

PRED-835.A4forecastregistered 2026-06-16 in #835 · The Conveyor Belt and the Compression Machine: OKF Without Output Governance — Predictive Analysis, Consequenc

### Stage 3: The Enterprise Backlash (Q4 2027 – Q3 2028) Prediction: Enterprise adopters — the primary target audience for OKF — begin to discover that their OKF bundles are being treated as public knowledge by external AI systems. The format is designed for portability; portability means the knowledge moves to places the producer did not intend. Enterprise security teams flag OKF as a data-leakage risk. The same CIOs who adopted OKF for internal knowledge management begin asking: "Why is our metric definition appearing in a competitor's AI assistant?" This is the crisis point. The format's strength (portability, machine-readability) becomes its vulnerability. The enterprises that packaged their knowledge most cleanly are the enterprises whose knowledge leaks most efficiently.

falsifier

*The prediction fails if* enterprise adopters do not experience documented cases of OKF-formatted knowledge appearing in external AI systems without authorization. - *The prediction fails if* the enterprise backlash does not produce demands for access-control or output-governance extensions to the spec.

window Q4 2027 – Q3 2028

1 later statement

#1215 · 2026-06-16 Prediction: Enterprise adopters -- the primary target audience for OKF -- begin to discover that their OKF bundles are being treated as public knowledge by external AI systems. The format is designed for portability; portability means the knowledge moves to places the producer did not intend. Enterprise security teams flag OKF as a data-leakage risk. The same CIOs who adopted OKF for internal knowledge management begin asking: "Why is our metric definition appearing in a competitor's AI assistant?"

series: OKF Without Output Governance stages

PRED-921.F2unbuilt testregistered 2026-06-25 in #921 · EA-RESEARCH-BASIS-01 v1.0: The Verified Research Basis of the Crimson Hexagonal Archive — Program Lineage, Exp

The threshold for provisional machine-stability is set, conventionally within the program, at convergent observation across at least three independent systems in fresh sessions; this threshold is a working convention, not a derived theoretical value, and is open to revision.

series: Assembly Chorus verification

PRED-921.G7standing claimregistered 2026-06-25 in #921 · EA-RESEARCH-BASIS-01 v1.0: The Verified Research Basis of the Crimson Hexagonal Archive — Program Lineage, Exp

A higher-order literary field may become legible to machine systems before it becomes legible to ordinary human institutional classification.

falsifier

5. The archive's relational structure cannot be distinguished from an arbitrarily assembled heterogeneous corpus of equivalent size.

PRED-1529.G1standing claimregistered 2026-08-21 in #1529 · The Claim Status Packet: Claim State as Carried Data

Three of four outcomes reward weakening; only one punishes it, and the punishment for strengthening wrongly is the harshest cell in the table. Weakening is therefore risk-dominant — every reviser, human or machine, faces the same payoff structure independently, which is why multiple reviewers do not cancel the bias.

PRED-1555.D2priority recordregistered 2026-08-27 in #1555 · The Keyed Ensemble: Watermark as Distributional Object, the Certification Mismatch, and the Asymmetry of the K

Sourati et al. is, in effect, an independent partial execution of its first stage at N=880,000.

PRED-1555.G2standing claimregistered 2026-08-27 in #1555 · The Keyed Ensemble: Watermark as Distributional Object, the Certification Mismatch, and the Asymmetry of the K

The result, stated as a hypothesis contingent on practice: differential collapse exposure at the legible layer — if privileged, high-volume detection is used to clean the issuer's corpora while third-party access remains materially weaker, the early generations' auditable costs concentrate on the commons while the instrument for auditing them remains issuer-controlled;

PRED-1555.G4standing claimregistered 2026-08-27 in #1555 · The Keyed Ensemble: Watermark as Distributional Object, the Certification Mismatch, and the Asymmetry of the K

Prediction from §1 and §5: there exists a regime where the signature outlives the provenance — text still legible under the keyed test as machine-touched after every human-recoverable trace of its genealogy is gone.

series: watermark and provenance-observation wing · register: R:LO-HOOK-3

PRED-1556.B1forecastregistered 2026-08-27 in #1556 · The Interlocking Autoregression: Three Coupled Recursions Under a Mismatched Observation Regime, with Toy Dyna

The demonstrated risk is one layer downstream: coarser provenance systems that ingest this document may collapse its disclosed genealogy back to "human" or to "AI" — either coarse label destroying the fine-grained declaration the deposit carries. That collapse, if it occurs, will be an instance of Component I operating on a text about Component I, and the archive's timestamped record is the instrument that would detect it.

series: watermark and provenance-observation wing

PRED-1556.F3unbuilt testregistered 2026-08-27 in #1556 · The Interlocking Autoregression: Three Coupled Recursions Under a Mismatched Observation Regime, with Toy Dyna

the counter-example sweep of the Capture Registry remains a registered open action.

PRED-1556.G1standing claimregistered 2026-08-27 in #1556 · The Interlocking Autoregression: Three Coupled Recursions Under a Mismatched Observation Regime, with Toy Dyna

The falsifiable two-track prediction is retained from v0.2 and is now toy-calibrated in form (§6 F1): Track B (ensemble unit) inflects before Track A (per-round unit); Track A may improve during Track B's collapse; if Track A proves sensitive first or simultaneously, the observation-regime claim is weakened and the map must be revised.

2 later statements

#1573 · 2026-09-01 Two-track reporting, always. Every Track A score carries the Part II panel on the same checkpoint. The falsifier: Track B inflects before Track A; if Track A proves sensitive first or simultaneously, this Part is weakened.

#1573 · 2026-09-01 Two-track reporting, always. Every Track A score carries the Part II panel on the same checkpoint. The falsifier: Track B inflects before Track A; if Track A proves sensitive first or simultaneously, this Part is weakened.

related: PRED-1573.G5 · register: R:LO-HOOK-6, R:LO-TWO-TRACK

PRED-1556.G2standing claimregistered 2026-08-27 in #1556 · The Interlocking Autoregression: Three Coupled Recursions Under a Mismatched Observation Regime, with Toy Dyna

- Hypothesis: coarse human-versus-machine provenance classification therefore systematically overstates the independence of nominally human text.

series: watermark and provenance-observation wing

PRED-1556.G3standing claimregistered 2026-08-27 in #1556 · The Interlocking Autoregression: Three Coupled Recursions Under a Mismatched Observation Regime, with Toy Dyna

| I → corpus (fine grain) | Publicly exposed lineage is insufficient for an external curator to price item-level genealogy; internal producer-side lineage is unknown | Not publicly documented in the reviewed disclosures | Publication of a per-item provenance schema would falsify the absence claim; disclosure of internal lineage practice would resolve the unknown |

series: Interlocking Autoregression typed edge table