Imagine selling your writing to an AI training company and receiving a check. You have no way to know if the amount is correct. The company runs an algorithm called Shapley value, a cooperative-game-theory method that calculates how much each data contributor is worth to the final model, and hands you a number. You cannot audit the inputs, cannot verify the computation, cannot independently confirm you were not shortchanged. You trust the operator or you do not. A paper published May 5, 2026, proposes to eliminate that trust entirely. It is called ZK-Value, authored by Zhaoyu Wang, Pingchuan Ma, and Zhantong Xue at arXiv:2605.03581v1. It is the first publicly documented system to apply zero-knowledge proofs specifically to Shapley-value data attribution at a scale that might actually work. And it arrives in a market where the problem has only gotten more urgent.

Data marketplaces are no longer niche. Hugging Face hosts thousands of datasets, many contributor-funded. Amadeus Lab, Databento, Numerai, and a dozen smaller platforms now operate revenue-share models where participants submit data and split proceeds based on some valuation method. Most use Shapley value because it is theoretically sound: it assigns each contributor credit proportional to their marginal contribution to model quality. But the operator alone knows the private data, the private model outputs, and the computation. An external auditor cannot independently verify the score without exposing everyone else's data. And participants cannot audit their own score without seeing the full dataset. So the operator is a single point of truth in a system that is supposed to be about fair distribution. This is the surveillance and trust problem that ZK-Value is designed to address.

Here is what ZK-Value actually does. Instead of the operator computing Shapley values in private and publishing the results, the system allows the operator to generate a cryptographic proof that the announced valuations are correct, without revealing the underlying data or model to anyone. A participant can then verify independently: given the proof and their own data submission, they can mathematically confirm that their attributed value is correct. No leakage of other contributors' data. No requirement to trust the operator's arithmetic. The proof itself is small enough to be broadcast alongside the payment distribution. The authors describe it as 'practical', the paper includes benchmarks, which means it is not purely theoretical. In the current near-zero-fee environment on Bitcoin (Mempool.space reports 2 satoshi per byte for fastest confirmation, 1 satoshi at 30 minutes as of block height 948,156), a marketplace operator could even anchor the Shapley proof to the blockchain or settle micro-payments over Lightning, making the entire audit trail public and unchangeable.

Why now? Three structural forces collide. First, the scale and economic stakes of data marketplaces have reached a point where trust failures have real consequences. Second, zero-knowledge proof technology has matured enough that proving Shapley-value calculations is no longer purely academic, the researchers have written optimized code. Third, the Bitcoin fee market has collapsed to near-floor, making it economically viable for the first time in years to settle small-value data payments on-chain or via Lightning without the fee exceeding the payment itself. A few years ago, proving a payment distribution would have cost more in fees than the payment was worth. Not anymore. The conditions are aligned.

Who benefits? Data contributors, the non-technical people selling their writing, code, or images to training datasets, suddenly have agency they did not have before. If a marketplace implements ZK-Value, they can verify they were paid fairly without hiring a lawyer or learning cryptography. They move from blind trust to mathematical certainty. The marketplace operator loses something different: the ability to silently adjust valuations, or to defend opaque scoring by claiming audits are too costly or data is too sensitive. The system is design-enforced fairness. The real winner, though, is the ecosystem thinking through trustless infrastructure: Fedimint already ships with three default modules (Bitcoin, Lightning, Chaumian Ecash) for private and trust-minimized payments. ZK-Value is exactly the kind of cryptographic primitive that fits into that model. You could build a data marketplace on Fedimint where payouts are private, instant, and provably fair, with zero involvement from a central operator for any of it. That is a different category of system than what exists today.

This is not a finished product. It is a research paper with no announced implementation, no live marketplace testing it, and no major publication yet covering it. The paper itself has not been peer-reviewed through a journal. But it addresses a real problem that affects an increasingly large and non-technical population, anyone who has ever considered selling their data to an AI company. And it does so using the same philosophical move that animated Nostr (decentralized social publishing without central servers), Fedimint (collective Bitcoin custody without a trusted custodian), and Lightning (payments without a banking intermediary). Move the point of control from an operator to mathematics. Let cryptography replace trust. The paper shows the math works. Whether it actually gets built and deployed, that is the open question.

Watch three things. First: Does a major tech publication cover ZK-Value in the next 72 hours? If The Block, Bitcoin Magazine, or a tier-1 cryptography outlet picks it up, the paper moves from academic to industry-relevant. Second: Does a Fedimint developer or ecash researcher engage with the Shapley-value angle? Fedimint is purpose-built for exactly this kind of provably-fair, trust-minimized payment system. If the Fedimint community sees the implication, a working prototype could appear fast. Third: Does the Bitcoin fee environment stay at this floor? At 2 sat/vB, a marketplace can settle Shapley proofs on-chain affordably. If fees spike, and they will eventually, the window for launching an economically viable blockchain-anchored proof system will have closed. The researchers built ZK-Value for this moment. The question is whether anyone acts before the conditions change.