2026 Manifest Passport Proof Layer Build vs Buy Start a Conversation
LegacyNext

LegacyNext · AI Rights Infrastructure

Prove what was licensed.
Settle it back to the work.

An execution layer for societies, labels and publishers entering the AI economy. It records which work it is, who may license it, what is permitted, exactly which works sit inside a training deal, and how the money is divided afterwards. Every statement is written to a blockchain, so both sides of a licence can prove the same facts years later.

What Changed in 2026 The Proof Layer
Core objectRights passport
Proof layerBlockchain-anchored
ScopeMaster & composition
OutcomeLine-item settlement

What changed

AI licensing became an evidence problem.

Within twelve months, courts, regulators and platforms all moved in the same direction. The operative question is now one of proof: what was permitted, by whom, and for exactly which works.

31 Jul 2026
The Regional Court of Munich I held in GEMA v Suno (42 O 763/25) that memorisation of protected works in model weights is a reproduction requiring authorisation, and that the text-and-data-mining exception does not cover it, even where the provider complies with the AI Act. Not final; appeal remains possible.
Nov 2025
The same court reached a comparable conclusion in GEMA v OpenAI (42 O 14139/24). Two rulings, one direction, and an AI provider now has its own reason to want hard evidence of what it licensed.
Since Aug 2025
Under Regulation (EU) 2024/1689, providers of general-purpose AI models must publish a sufficiently detailed summary of training content and operate a copyright policy respecting rights reservations. That summary sits at corpus level. Settling and auditing a licence needs work-level data.
13 Aug 2026
Spotify licensed Kobalt for fan-made AI covers and remixes, following Universal Music Group and Merlin. Repertoire is now used two ways, to train models and to let users generate new versions, and each is a separate permission on the same work.
Consequence
A reservation against text-and-data mining is worth little unless it is machine-readable and findable per work, and the same holds for its mirror image, positive consent under stated conditions. Both have to live somewhere a counterparty can verify.

The gap

The break sits between rights and money.

Rights exist. Money arrives. What fails is everything in between, and it fails in the same six places at every organisation we have looked at.

Identity
The same work carries different identifiers across systems, and recording and composition are poorly connected. One track ends up with several identities, none of them authoritative, and every missing link raises the odds of a payment that cannot be attributed to any work.
Mandate
Who is currently entitled to permit a specific use is rarely recorded in a form a machine can read. Ownership answers a different question.
Consent
Permission is captured as one blanket yes or no, while the market prices training, fine-tuning, voice, covers, remixes and synthetic performance separately.
No usage event
Streaming counts plays. A corpus licence pays a lump sum and counts nothing, so the split becomes a policy choice rather than a measurement, which is a liability question for a board.

First product

The verified corpus manifest.

Every training licence ships with a signed, hashed list of exactly the works inside the licensed dataset. It is simultaneously the proof of scope for the licensee and the starting point for allocation for the rightsholder. Neither side has to trust the other's database.

What the manifest carries

  • Recording and composition IDs, alongside ISRC, ISWC and existing identifiers
  • Version relationships: remixes, remasters, alternate takes
  • Rightsholders and the mandate as it stood on the licence date
  • Consent per use type, territory and term
  • Hash and signature, anchored so tampering is detectable

What it lets you prove

  • Scope: exactly which works were inside the deal, and which were not
  • Authority: who was entitled to grant the permission, at that moment
  • Allocation: a defensible split from a known set instead of a discretionary one
  • Defence: a licensee can show a court what it licensed and from whom

It starts from facts that are already settled: a known dataset, known works, known rights, a known payment. That is far easier to defend than a claim about how much any single training work influenced a given output.

The record

A rights passport in four layers.

Recording and composition stay separate objects with an explicit, verifiable link between them, because the rights, the holders and the mandates genuinely differ. The value sits in the link.

Identity

Recording and composition identified reliably, with versions and derivatives attached to the source record rather than floating free.

Title & mandate

Who holds which right, in which territory, for which period, and which organisation is entitled to license it today.

Consent

Machine-readable permission or refusal, per use type and per counterparty, alongside the reservation that mirrors it.

Settlement & proof

Every allocation tied to a licence, a period, a money flow, a passport version and a split key, provable line by line.

Master object

ISRC, recording owner or producer, performers, neighbouring rights, active mandates, recording versions.

Composition object

ISWC, composers, lyricists, publishers, society mandates and splits, linked verifiably to the recording.

The proof layer

Why this runs on a blockchain.

A rights record is only worth something when more than one party will accept it. That is the whole reason the proof layer exists, and it is the part of LegacyNext that cannot be reproduced by writing a better database.

Tamper-evident by construction

Who declared what, under whose authority, for which scope, at which moment, and when it changed. The history cannot be rewritten quietly, which is exactly what an auditor, a regulator and a counterparty each need.

One token per work

Each work carries a unique, non-fungible token with a single continuous history. Issuing is limited to the mandate holder and bound to the work's identifiers, so a duplicate registration is refused at source rather than argued about later, and a competing claim runs through a documented conflict procedure instead of settling in whichever system is consulted first.

Cross-party evidence

The licensee proves scope, the rightsholder proves entitlement, and both read the same anchored record. No party has to accept the other's internal system as the source of truth.

A record, not a tradable asset

The token carries a rights record and represents no financial claim on revenue. It is unique, non-fungible and not offered on a secondary market, which is the ground on which counsel assesses it under MiCA. Where a structure does carry financial exposure, that is arranged separately under the applicable regime.

Standards stay

DDEX remains the exchange language, ISRC and ISWC the identifiers, societies the licensors. The record feeds those systems, which is where earlier global-database attempts failed.

The precise claim

The blockchain registers and proves; it does not transfer copyright. Dutch law requires a deed under Article 2 of the Copyright Act, and the deed stays. Being exact here is what makes the rest survive a legal review.

What the chain delivers is a verifiable rights record. Ownership keeps its own legal route, and holding that line is what carries the rest through a legal review.

Build vs buy

The code was never the expensive part.

A rights database is the cheap half of this problem. The cost sits in chain-of-title semantics, mandate modelling per territory, DDEX conformance and the discipline to keep a record provable while it changes, and that is where in-house builds run aground in year two.

The real cost
Software is a minority of the work. What consumes the years is chain-of-title semantics, mandate modelling per territory and per rights category, DDEX conformance, and the migrations that follow every schema decision. A wrong split is not a bug, it is a liability.
Operating, not building
The hard part starts after launch: key management, versioning a record that stays provable across changes, reconstructing a payment line by line for a regulator years later. Generated code comes with no operating discipline.
Evidence needs neutrality
What you build yourself is your database. A licensee will not accept the licensor's own system as proof, and the reverse holds equally. A record only becomes evidence when neither party controls it alone, which is why whoever administers mandate and consent should hold no rights position of its own.
Moving ground
Standards, licence models and volumes are all still shifting, and the Munich rulings can still go on appeal. Schema decisions taken this year will need revisiting as that settles. Buying the execution layer means someone else owns that migration.

The boundary

You remain the licensor.

The role a rights organisation must occupy in the AI economy splits cleanly in two. LegacyNext offers itself for one half only, and that is the design principle rather than a concession.

Authority: stays with you

  • Mandate: which AI rights are collectively administered, per use type
  • Licensing: granting the licence, setting the terms and the price
  • Governance: minimum conditions on transparency, audit, remuneration, creator autonomy
  • Consent decisions: the legal act of permitting or refusing

Execution: where we fit

  • Repertoire identification: work, master, composition, rightsholder data
  • Corpus administration: signed work-level manifests per licence
  • Consent administration: recording the decision machine-readably
  • Recognition signals: connecting third-party attribution to rights data
  • Settlement: matching, waterfall, reconciliation, payment
  • Interoperability: DDEX, ISRC, ISWC, existing standards

Who gets what

One layer, four counterparties.

The same record answers a different question depending on which side of the table you sit, which is precisely why it has to be shared infrastructure rather than anyone's internal system.

01

Collecting societies

Work-by-work consent and reservation, administration of licensed corpora, explainable allocation of lump sums, and an audit trail that holds when a board asks how a split was decided. Relevance early in the chain, at mandate and clearance.

02

Labels

Provable rights position on the master side, clearance data ready before an AI partner asks, settlement that reaches performers and producers per line, and declared provenance per work as catalogues fill with material of uncertain origin.

03

Publishers

Composition-side clearance where it is hardest: multi-writer works, split confirmation, and derivative passports for AI covers and remixes tied back to the parent work. The Spotify deals land here first, and so does the administrative load.

04

AI providers & auditors

Proof of licensed corpus, current mandate and consent status, and reporting that maps onto the AI Act's training-content obligations, with money flows and split keys reconstructable line by line years later.

Boundaries

Where the limits sit.

Stating the limits early is what makes the rest credible in a room with lawyers in it.

Licensing
LegacyNext registers and executes within the authority it is given. It does not license repertoire that is already mandated elsewhere.
Attribution
Similarity is an approximation of resemblance, presented as a signal with the causal question left open. Recognition technology comes from specialist providers and is connected to rights data, not reinvented here.
Identifiers
A verifiable record per work, built on ISRC, ISWC and DDEX, which all keep their function. The scope is one organisation's repertoire at a time, and the Global Repertoire Database's collapse in 2014 is the reason for that discipline.
Legal transfer
The passport records and evidences. A deed or other required legal basis remains necessary and untouched.

Next step

Start with one licence.

A first deployment is deliberately small and predictable: one hundred to a thousand works, master and composition linked, mandate status established, consent recorded per use type, a single licence or licensee, a signed manifest, one payment, and automatic allocation afterwards.

That proves four things at once: rights data links reliably at work level, mandate and consent become machine-readable, a corpus can be delimited objectively in advance, and the money divides explainably afterwards.

A sixty-minute conversation about your current AI exposure is enough to tell whether this fits. If there is a basis to continue, we agree the next step together.

Start a Conversation Technical Deep Dive