A protocol for anchoring evidence of use, expressing license terms as structured records, and settling royalties against revenue that was sourced rather than self reported. Records are secured by attestors who post capital and lose it when they are wrong.
Brand licensing is a large, old, and unusually badly instrumented market. Rights are granted in prose, recorded in scattered documents, sublicensed without a shared registry, and settled against sales figures the paying party calculates about itself. Licensors budget for forensic accountants as a routine cost of doing business, and a question as basic as who currently holds what rights in which territory frequently cannot be answered in a single query.
Probant proposes a shared record layer with three properties. Anchoring evidence of use is free and permissionless, so the registry densifies without a sales motion. License terms are expressed as structured, machine checkable records rather than prose, so a sublicense that exceeds its parent term is a constraint violation rather than a clause somebody has to notice. And the records that matter are secured by attestors who post a bond, face a challenge window, and are slashed when a challenge succeeds.
The network token exists to be that bond. It is not a payment rail, not a fee token, and not a governance ornament. Rights holders transact in dollars and are never required to hold it.
Scope statement. Probant records evidence and administers arrangements. It does not create, replace, register, or adjudicate any intellectual property right. A record in Probant is evidence a tribunal may weigh. It is not a determination.
Not the bonding. Bonded assertion with a challenge window is proven infrastructure, running at scale in optimistic oracles and staked arbitration today. What is new is the domain, and one property that domain supplies for free.
Every bonded system faces the same unsolved problem: slashing deters only if misbehavior is detected and reported, and protocols generally have to invent a reason for anyone to go looking. Brand rights arrive with that reason already paid for. Acquirers in diligence, licensees in territory disputes, licensors who suspect underpayment, and litigators preparing oppositions all have their own money riding on finding a defective record. The audit incentive is external and pre-existing.
A standard license obligates the licensee to report sales periodically and remit a percentage. The licensor's visibility into whether that figure is correct comes from an audit right it exercises rarely, late, and at its own expense. Royalty audit is a standing professional service line precisely because the reported number and the correct number differ often enough to fund an industry.
Most of the difference is not fraud. It is currency conversion applied at the wrong date, returns and allowances handled inconsistently, channel mix that falls outside the definition of net sales, and sublicensees who never received the terms they were supposed to be bound by. These are record keeping failures, and record keeping failures are the class of problem a shared ledger actually solves.
Rights move. They are assigned in acquisitions, split by territory, carved by product class, and sublicensed downstream, each step documented in a separate instrument held by a different party. Reconstructing the current state means collecting documents from counterparties who have no obligation to respond quickly. In diligence this becomes a schedule of exceptions. In litigation it becomes discovery.
Rights in a mark depend on use in commerce. Defending, renewing, or opposing frequently turns on proving that a mark was in use, in a form, on a date. Today that proof is assembled retroactively from marketing archives and invoices. The World Intellectual Property Organization has noted that distributed ledgers are well suited to recording timestamped evidence of use and rights management information, while being unable to replace registration itself. That boundary defines the scope of this protocol.
Rights and royalty platforms are capable and widely deployed. Rightsline alone reports more than 135 million IP assets under management across entertainment, gaming, publishing and life sciences. They replace spreadsheets with a central repository and compute royalties across thousands of contractual exceptions.
Every one of them is a system of record for a single licensor. That is the correct architecture for internal operations and the wrong one for a problem between counterparties. The licensee's reported sales still enter the system as a number the licensee typed, and the licensee's own system holds a different copy of the same agreement. Nothing in a single sided database can make two parties who do not trust each other agree on a fact.
The mechanism proposed here is not new. Bonded assertion with a challenge window and slashing is a proven pattern with several live implementations at scale. What is new is pointing it at brand rights, and the honest reason to catalog the prior art is that the closest historical analog largely failed.
Token curated registries, proposed in 2017, are almost exactly this idea: a list whose entries are staked, challengeable, and voted on. Nearly all of them failed. The documented reasons are specific and worth naming rather than eliding.
Why this design is not a token curated registry
The claims are objective. Was this specimen in use on this date. Does this sublicense grant a term exceeding its parent. Does this reported revenue match the connected source. These have correct answers a challenger can demonstrate, which is the opposite of asking whether a newspaper is reputable. Schelling point pathology arises when there is no fact to converge on. Here there is one.
Challengers are subsidized from outside the system. In a token curated registry the only reward for challenging is the bond. Here, an acquirer in diligence, a competing licensee, or an underpaid licensor already has millions of dollars of their own reason to find a bad record. The protocol does not have to manufacture the incentive to audit it. a16z's analysis of slashing economics treats incentivizing detection and reporting as the unresolved half of the problem. This domain resolves it by accident.
The registry has value without the token. Free anchoring produces a dense, useful record whether or not anyone ever stakes. Bonding secures the subset of claims where money turns on the answer. A token curated registry had nothing underneath the token.
Genesis distribution is deliberately not concentrated. Team and investors together hold 31%, disclosed, vested, and cliffed, against adChain's 40% to two entities.
Holders never decide a dispute. Governance covers parameters. Individual disputes escalate to a domain qualified panel, which closes off the exact mechanism that produced the canonical failure.
Story Protocol operates a chain for programmable IP licensing across creative works: music, art, writing, software, and AI training data, with modules for licensing, royalty distribution between originals and derivatives, disputes, grouping and metadata. It does not address registered trademarks, royalty reporting from real world sales, or sublicense chains inside commercial brand licensing programs. Reporting also notes that onboarding large IP holders has been slow. It is a neighbor rather than a rival, and the two systems could reasonably interoperate.
Probant deploys on Base. Contracts are minimal by intent, because the durable asset is the record rather than the logic, and logic that holds value is logic that must be audited.
MarkRegistry anchor(bytes32 digest, bytes32 schemaId, bytes meta)
transferTitle(bytes32 markId, address to)
linkRegistration(bytes32 markId, string office, string serial)
LicenseRegistry grant(bytes32 markId, LicenseTerms terms) returns (bytes32)
sublicense(bytes32 parentId, LicenseTerms terms)
// reverts when terms exceed the parent envelope
AttestationHub attest(bytes32 subject, bytes32 claimHash, uint256 bond)
challenge(bytes32 attestationId, uint256 bond)
resolve(bytes32 disputeId) // escalates to arbitration
BondVault bond(uint256 amount) / unbond(uint256 amount)
// unbonding delay covers the open challenge window
FeeRouter collect(uint256 usdc)
// splits to attestor yield, buyback and burn, operations
SettlementBook report(bytes32 licenseId, RevenueReport r, bytes sourceProof)
settle(bytes32 licenseId) // pays royalty in USDCOnchain: digests, attestations, bonds, license term structs, settlement events, and the resolution of disputes. Offchain: every document, every specimen image, every underlying sales record. The protocol never takes custody of a customer document and never publishes commercial terms the parties have not chosen to make public. A license record may be published in full, published as a redacted envelope carrying only the fields a third party needs to check for conflict, or kept private with only its digest anchored.
Anchored digests are meaningless without the documents that produce them, which is a feature for confidentiality and a liability for continuity. Rights holders retain the originals. Optional encrypted escrow with a third party custodian is available for parties who want the record to survive their own document management.
Every claim resolves to one of a small number of schemas. Keeping the set small is what makes challenges tractable, because a challenger must be able to state precisely what is false.
| Schema | Claim | How a challenge is proved | Bond band |
|---|---|---|---|
| use.specimen.v1 | This specimen depicts the mark in commercial use on or before this date | Producing the specimen with a provably later origin, or showing the digest does not match | Low |
| title.chain.v1 | Title passed from party A to party B on this date under this instrument | Producing a conflicting recorded assignment | High |
| license.envelope.v1 | These terms are the operative grant for this license | Producing the executed instrument with differing terms | High |
| sublicense.valid.v1 | This sublicense falls within its parent envelope | Arithmetic on the parent record. Largely machine checkable | Medium |
| revenue.sourced.v1 | This revenue figure came from the connected source for this period | Re-querying the connector, or showing the signature does not verify | Medium |
Bonds scale with the value at stake in the underlying arrangement rather than being flat. A flat bond either over-collateralizes trivial claims or under-collateralizes consequential ones. The governing inequality is the standard one: the cost of a false attestation, meaning the bond forfeited multiplied by the probability of detection, must exceed the profit available from making it. Because detection here is externally funded, the probability term is unusually high, which permits smaller bonds than a comparable system would need.
Ten billion tokens, fixed at genesis, no mint function afterward. Allocation and the eight year emission schedule are published in full on the tokenomics page. Three design rules govern the system, and each one exists because a named predecessor got it wrong.
Rights holders pay in USDC for the licensing product. The fee router splits that revenue three ways: a share to bonded attestors as dollar yield, a share to buy the token on market and burn it, and the remainder to operations. The router never holds customer float, and royalties settle party to party.
Token demand therefore comes from wanting the dollar yield, which requires bonding, which requires acquisition. Demand scales with attested value rather than with marketing.
At low network revenue the emission schedule exceeds the burn and the token is inflationary. There is no arrangement of parameters that avoids this, and the supply model on the tokenomics page lets anyone verify it by dragging revenue toward zero. Publishing that rather than concealing it is the position: a token whose economics only work at scale should say so before it has scale.
Governance is deliberately narrow at launch and expands as the network demonstrates it can handle the responsibility. Holders govern fee split parameters, bond band calibration, schema additions, and treasury allocation. Holders do not govern the outcome of individual disputes, which is the failure mode that produced adChain's rejection of the New York Times. Dispute resolution escalates to an arbitration layer with domain qualified panels rather than to a plebiscite of token holders.
A time locked multisig retains emergency pause authority over the bond vault and settlement contracts for the first twenty four months. This is a centralization the protocol admits to rather than obscures, and its expiry is written into the contract rather than promised in a blog post.
The SEC proposed Regulation Crypto Assets on 18 August 2026, published in the Federal Register on 21 August, with a sixty day comment period. It contemplates a startup exemption of five million dollars over four years with narrative disclosure and no accredited investor requirement, two larger fundraising tiers, and a safe harbor by which an investment contract ceases to exist once the issuer certifies that essential managerial efforts have completed or permanently ceased.
Our position, stated plainly. No token exists. No sale is open. No price has been offered. We intend to distribute under the startup exemption once rules are final, and we run a non transferable contribution points program in the interim. We are incorporated in the United States and satisfy the US nexus test that projects domiciled offshore during the last cycle now fail.
Two risks we do not minimize. The safe harbor depends on issuer self certification and the Commission may later challenge whether managerial efforts genuinely ceased. And the proposal may change materially before adoption, in which case this section changes with it.
Version 0.1, 25 August 2026. This document describes a system under construction. Nothing in it is an offer to sell or a solicitation to buy any security, none of the figures are projections or promises of value, and none of it is legal advice. Probant Protocol and PRBT are names of a system under construction.