Mechanism

Creator attribution in this disclosure is not a label embedded in content. It is a computation performed over three structures the platform already maintains: the variance-derived unique identifier (UID) of a content artifact, the alias record a creator registers for that UID, and the consultation event log produced when a generative model queries a reference artifact. The UID is derived deterministically from the artifact's internal variance and structural features rather than from its storage location, file name, cryptographic key, or transmission metadata, so the identity that attribution attaches to follows the content's structure rather than its packaging. Attribution and compensation are then computed from records that reference that UID, not from inspection of model weights or training logs.

The architecture operationalizes a single principle: content creator compensation for generative AI use of training data should be computable from verifiable structural proximity measurements rather than from approximations of training-data influence derived from model weight analysis. To make that computable, the disclosure requires three pre-conditions. The training-data artifact must have a registered UID in the anchor network. The creator must have registered an alias for that UID with a compensation routing field specifying a payment address and a compensation schedule. And the generative model must operate within a governed execution environment that logs consultation events when the model queries reference artifacts through retrieval-augmented generation or structured neighborhood resolution during inference.

The Alias Record and Compensation Routing Field

Attribution attaches through the alias. An alias is a human-readable symbolic identifier registered in association with a variance-derived UID, allowing content creators, institutional publishers, platform operators, and agents to assign a persistent symbolic reference to a content object without replacing or conflating the object's deterministic, variance-derived identity. Aliases are treated as mutable, scoped references governed independently of the UID itself and resolved through slope-band-scoped anchor consensus.

For attribution, the alias record carries more than a name. A content artifact with a registered alias has an alias record comprising the alias string, the variance-derived UID, a compensation routing field encoding a payment address, a compensation schedule reference, and a jurisdictional scope. The compensation routing field is what makes a creator addressable: it names where an obligation should be credited and under which schedule and jurisdiction. The payment address may designate a bank account routing number, a digital payment service identifier, a blockchain wallet address, an internal ledger account within a platform payment system, or any other machine-resolvable payment endpoint. The architecture is payment-infrastructure-agnostic: the routing field specifies the destination and the schedule specifies the amount, and the payment routing module resolves the destination format and executes the credit through the appropriate payment interface.

Consultation Events

The signal that drives attribution is the consultation event. For each generation event that consults a reference artifact through retrieval-augmented generation or structured neighborhood resolution, the consultation event logger records a deterministic consultation record comprising the variance-derived UID of the consulted artifact, the governing policy version, the variance proximity score between the consulted artifact and the generated output, and a timestamp. These records exist because consultation happens at a governed boundary, not because the model's internals are inspected.

This relocates attribution from an inference problem to a logging problem. Rather than reverse-engineering which training data influenced an output by analyzing model weights, the platform records the consultation at the moment it occurs and ties it to a UID that is itself derived from the consulted artifact's structure. When generation consults reference artifacts within policy-defined similarity neighborhoods, those consultation events are logged as computable attribution events from which compensation obligations may be derived under policy-declared schedules. Creator payment becomes computable from consultation event logs rather than from approximations of training-data influence.

Computing the Attribution Weight and Payment Obligation

The attribution computation module aggregates consultation events per consulted artifact and computes an attribution weight as the product of consultation frequency and mean variance proximity score. An artifact that is consulted often, and whose structural proximity to the generated outputs is high, accumulates a larger weight than one consulted rarely or only at the margins of a similarity neighborhood. The weight is therefore a function of two measured quantities the platform already records: how many consultation events name the artifact, and how close those consultations were in variance space.

The compensation computation module multiplies the attribution weight by the schedule rate drawn from the compensation schedule reference and produces a payment obligation record. The payment routing module then credits the computed obligation to the payment address associated with the consulted artifact's alias record, under the declared jurisdictional scope. The compensation schedule reference is a versioned, machine-evaluable, cryptographically signed document specifying the rate structure under which attribution weights translate to payment obligations. Schedule formats include per-consultation flat rates, proximity-weighted rates in which higher cosine similarity scores produce proportionally higher payments, category-specific rates, and jurisdiction-specific rates. Schedule versioning ensures compensation computations are reproducible and auditable from the schedule version in effect at the consultation event.

Derivative and Composite Attribution

Attribution also operates across derivation, not only across generative consultation. The platform supports composite content attribution by enabling each UID to participate in a directed lineage graph that may contain more than one parent. Content artifacts generated through combination, transformation, remixing, or derivation from multiple prior artifacts may be registered under a new UID linked to all contributing source UIDs, with lineage edge weights reflecting the degree of variance inheritance from each contributor.

Weighted parent attribution assigns a contribution weight to each confirmed parent UID proportional to its cosine similarity with the derivative UID. These weights are recorded as edge annotations in the directed lineage graph. The disclosure is explicit that they do not constitute legal determinations of authorship or ownership; they serve as structural signals that may inform licensing, attribution display, policy inheritance, and governance enforcement by downstream systems. Per-quadrant comparison supports spatially resolved attribution: a derivative that modifies only one spatial region of a source exhibits quadrant similarity scores near 1.0 for unchanged regions and reduced scores for the modified region, so attribution can be localized even when the global similarity score remains high.

Auditability and Adversarial Resistance

The compensation audit log maintains an append-only record of payment obligation records, comprising the consulted artifact UID, the computed attribution weight, the schedule version applied, the payment obligation amount, the payment address credited, the jurisdictional scope, and a timestamp. The log is verifiable against the consultation event log: any party may confirm that payment obligations are consistent with the corresponding consultation events, the compensation schedules in effect at the relevant times, and the computed attribution weights, without requiring access to model weights, training logs, or proprietary platform data.

Attempts to evade attribution are detectable through the same structural measurements. A composite content artifact is considered fully auditable: any party with access to the governing anchor cluster may trace a composite UID to its contributing parents, inspect mutation metadata, evaluate policy lineage, and verify anchor quorum endorsements. Adversarial recombinations, in which a party attempts to register a derivative that suppresses attribution to one or more source UIDs by manipulating variance features, are detectable because the slope profile of the composite UID exhibits measurable divergence from the weighted variance combination of its declared parents. Anchors may flag such entries and subject them to heightened governance review. The delegation policy evaluator enforces ownership delegation constraints at registration and mutation time, denying registration when, for example, an unauthorized agent attempts to register a derivative UID from a policy-protected source.

Prior-Art Distinction

Watermarking and metadata tagging approaches embed identity signals in the content stream or a sidecar record. Watermarks are removable through transcoding, cropping, or generative reconstruction, and metadata records are decoupled from content structure and require persistent external storage. The mechanism described here attaches attribution to a variance-derived UID computed from the content's own structure and computes compensation from logged consultation events, so nothing severable need be embedded in the artifact for attribution to be computable.

Blockchain-based asset registration systems anchor ownership records to distributed ledgers through hash-based proofs of existence, binding identity to key-pair ownership rather than content structure, and cannot detect semantic similarity between related or derivative content objects. The architecture here binds attribution to structural identity and weighted lineage rather than to key ownership, and remains payment-infrastructure-agnostic: a blockchain wallet is merely one of several admissible payment endpoints in the compensation routing field, not a requirement of the attribution computation.

Approaches that estimate creator compensation by analyzing model weights or approximating training-data influence require access to the model's internals and yield non-reproducible estimates. The disclosure replaces that with attribution weights computed from consultation frequency and mean variance proximity, and payment obligations reproducible from versioned compensation schedules, all verifiable without access to model weights or training logs.

Disclosure Scope

The disclosure of creator attribution covers the computation of attribution and compensation from a creator's alias record, the consultation event log, and the variance-derived unique identifier of a content artifact. It covers the alias record's compensation routing field encoding a payment address, compensation schedule reference, and jurisdictional scope; the consultation record comprising the consulted artifact UID, governing policy version, variance proximity score, and timestamp; the attribution weight computed as the product of consultation frequency and mean variance proximity score; the payment obligation record produced by multiplying the attribution weight by the schedule rate; the payment routing of that obligation to the address in the alias record; and the append-only compensation audit log verifiable against the consultation event log. It covers weighted parent attribution over multi-root lineage graphs, in which contribution weights are assigned proportional to cosine similarity and recorded as edge annotations without constituting legal determinations of authorship, and the spatially resolved quadrant-level attribution and adversarial-recombination detection that accompany it.

The disclosure does not depend on any particular payment infrastructure, content modality, or compensation schedule format. The payment address may resolve to any machine-resolvable payment endpoint; the schedule may specify per-consultation flat, proximity-weighted, category-specific, or jurisdiction-specific rates; and the consulted artifact may be of any modality for which a variance-derived UID can be computed. Substitution of these implementation details does not exit the disclosure. This subject matter is disclosed in PCT International Application No. PCT/US26/28630.