BIS Tests the XRP Ledger to Lock Official Statistics

BIS researchers published a 56-page working paper on September 2, 2026 that timestamps official statistics on the XRP Ledger. The prototype is a public notary for files that central banks already share, not a payments product and not a live rollout.

The authors wrote that SDMX, the format used to move those files, still cannot prove a downloaded set came from the issuer or that nobody changed it. They named third-party platforms and artificial intelligence systems as the places where that break now happens.

A Trust Gap Inside the SDMX Standard

Official statistics feed interest-rate decisions, bank stress work, and the numbers the public is asked to believe. SDMX already standardises how those files move among international bodies. It does not, on its own, let a user check origin or integrity after a file leaves the publisher.

The working paper on verifiable official statistics puts that gap in the open. Mario Rusev of d-fine Austria GmbH wrote it with four BIS staff: Rafael Schmidt, head of MED IT; Edward Lambe, head of MED data engineering; Christian Schmieder, head of operations; and Glenn Philip Tice, a consultant and product manager in the Monetary and Economic Department.

They point to a separate cost. An NBER working paper by Nicholas Bloom, Erica L. Groshen, Duncan Hobbs and Michael R. Strain, issued in April 2026 as working paper 35135, studies market uncertainty after the August 1, 2025 dismissal of U.S. Bureau of Labor Statistics Commissioner Erika McEntarfer. The BIS text cites that event-study as evidence that lost trust in official figures carries a real price.

Bloom and his co-authors estimate economic benefits of trust in official statistics of about $25 for every $1 spent on the agency’s budget. That figure is about the BLS, not about any ledger. It is the reason a statistics shop would pay for an independent check at all.

The views in working paper 1374 are the authors’ own and do not necessarily reflect the BIS or its member central banks. The paper also says references to firms and projects are illustrative, not a finding on anyone’s legal status.

What Working Paper 1374 Built on the XRP Ledger

Each SDMX file gets a cryptographic fingerprint. Where needed, each time series inside the file gets one too. Those fingerprints are rolled into a single Merkle root, then written to the XRP Ledger, where the entry is timestamped and cannot be edited later.

The published file carries a digitally signed credential tied to the publisher. A user can confirm who released the data and whether the figures still match, from the file plus one ledger lookup, with no call back to the publisher’s own servers. Only fingerprints go on-chain. The underlying series stay off it, so confidentiality holds.

On the prototype, publishing took 3 to 5 seconds and a check took 1 to 2 seconds, under controlled test conditions. The authors call those times fast enough for interactive use and for automated systems that consume data in real time. They also say the figures describe a proof of concept, not a hardened production system.

THE PROTOTYPE UNDER TEST

On-chain fees are described as negligible once modest batches are used, because a single ledger entry can cover thousands of data sets. Ordinary processing and storage then dominate the bill. The practical trade-off is batching for cost against the delay it adds to an urgent release; the paper solves for an economically optimal batch size.

The Proof Goes Into a Payment Memo

The researchers did not put a smart contract in the path. They used the ledger as a public bulletin board. A root is written into a Payment transaction memo, which is enough to timestamp the batch and make later edits visible.

That choice is why this is a notary job. It does not move value with XRP as a bridge asset, and even a scaled version would add little transaction volume. The fee exists only so the ledger will accept the record.

The BIS Open Tech write-up describes a cryptographic trust layer for SDMX that plugs into registries and dissemination workflows already in use. Core maintainers are listed as the Bank for International Settlements and d-fine. The same page names DeFi applications and AI-agent verification pipelines as consumers that can finish a check in 1 to 2 seconds without depending on the publisher’s backend.

HOW A FILE GETS ANCHORED

1. Normalise: Standard SDMX-ML payloads are canonicalised with W3C C14N 1.1 so the same file always hashes the same way.
2. Fingerprint: SHA3-512 is applied at whole-file level and, where needed, per series.
3. Bundle: The hashes are aggregated into a domain-separated, length-prefixed Merkle tree.
4. Anchor: The root is written to XRPL DevNet in a single transaction.
5. Carry proof: The published file is rewritten so it holds the root, the ordered leaves, and a W3C Verifiable Credential signed under the publisher’s XRPL key.

Anyone who later receives that file can re-derive the root and check the publisher’s identity from the file plus one ledger lookup. The GitHub readme for the open-source reference implementation calls it a proof-of-concept application that publishes and validates SDMX messages on XRPL DevNet, then writes the transaction reference back into the SDMX Source header.

Eight Agencies Already Share Data This Way

SDMX is not a BIS house format. It is the ISO standard for statistical data exchange (ISO 17369), sponsored by eight international organisations that already push the same files through the same pipes.

SDMX SPONSOR ORGANISATIONS

• BIS: Bank for International Settlements, which also ran this prototype.
• ECB: European Central Bank.
• Eurostat: Statistical Office of the European Union.
• IMF: International Monetary Fund.
• OECD: Organisation for Economic Co-operation and Development.
• UN: United Nations Statistical Division.
• World Bank: World Bank Group.
• ILO: International Labour Organization.

The first version of the technical standard was finalised in 2004. The first global SDMX data exchanges, in 2013, covered national accounts, balance of payments, and foreign direct investment. Version 3.1 of the technical specifications was released in May 2025.

If those sponsors ever attach a public fingerprint to the files they already send each other, the check travels with the data. A portal, a nowcasting desk, or a model that pulls a CPI or GDP series could test the file before it is used. That is the demand side the paper is written for, and it is a statistics-office problem long before it is a token problem.

The Paper Names AI Agents as the Next User

The BIS account posted the paper on September 2, 2026 as a question about downloaded files, not as a partnership with any token issuer.

Trust in official statistics underpins evidence-based policy. But how can a user know a data set they downloaded is the one the publisher released? Our new paper proposes an answer.

Bank for International Settlements, official account, September 2, 2026

https://x.com/BIS_org/status/2095120327495782466

The HTML summary is more specific. SDMX, it says, has no way for users to check that a data set really came from the issuing institution and has not been altered since publication, for example by third-party platforms or artificial intelligence systems. The abstract then lists automated verification by AI agents as a foundation the prototype is meant to support, alongside zero-knowledge proofs as a later extension.

That is the second effect. Models already ingest official series at a speed no human desk can audit line by line. A hash that travels with the file gives those models a machine check before the numbers enter a forecast, a trading tool, or a public dashboard. The SDMX site itself now lists “AI Ready” among its aims, which is why a statistics standard and a public ledger showed up in the same paper.

The same architecture, the authors write, can be extended to other structured reporting formats such as XBRL. The method is data-format-agnostic once a file can be canonicalised and hashed.

DevNet Results and the Production Gap

Every timed run used XRPL DevNet, not mainnet. DevNet shares the live network’s transaction format and ledger-close cadence, which is why the authors treat the 3 to 5 second publication window as informative. It is still a developer network. The GitHub project is labelled as research, and the paper’s findings section repeats that the timings are not a production service.

WHAT WE KNOW

• The paper: Working paper 1374 was published on September 2, 2026, with code under Apache 2.0.
• The test bed: Anchoring ran on XRPL DevNet, using Payment memos and SHA3-512 fingerprints.
• The scope: Only fingerprints go on-chain; the series stay off it.
• The disclaimer: The views are the authors’ and are not those of the BIS or its member central banks.

WHAT IS UNCONFIRMED

• Production use: No decision that BIS or any sponsor will run this on mainnet.
• Payments role: No claim that XRP is being used for central-bank settlement in this work.
• Other chains: The write-up is blockchain-based; this prototype is XRPL-specific, and no production bake-off is published here.

Limitations in the paper treat operational stability, cost, governance, and record retention as work still to do. A public timestamp is cheap and available. A statistics agency still has to decide who holds the publisher keys, how long the on-chain record must last, and what happens when a release is revised.

The honest read is optionality on a public integrity layer for official files. It is not a catalyst on its own, and it is not evidence that the BIS has adopted the XRP Ledger as infrastructure. The code is out. The tests are on a developer network. Statistical offices and the models that drink their files are the parties with something to gain if the method ever leaves the lab.

Frequently Asked Questions

What Is SDMX and Who Sponsors It?

SDMX, Statistical Data and Metadata eXchange, is ISO 17369, a standard for describing and sharing official statistics. Eight sponsors maintain it: BIS, the ECB, Eurostat, the IMF, the OECD, the United Nations, the World Bank, and the ILO. The first global exchanges in 2013 covered national accounts, balance of payments, and foreign direct investment, which is why a fingerprint method aimed at SDMX files would, if adopted, sit on pipes those bodies already use.

Did the BIS Adopt the XRP Ledger for Official Statistics?

No. Working paper 1374 is a research proof of concept on XRPL DevNet, and the authors state that their views do not necessarily reflect the BIS or its member central banks. The reference implementation is published as open source under Apache 2.0 so other statistical shops can inspect it; that is not the same as a production mandate or a payments contract.

Does the System Put GDP or Inflation Figures on a Public Blockchain?

It does not. Only fingerprints, a hash of the canonicalised file and, where used, of individual series, are written on-chain. Per-series hashing means a user can check one series inside a larger file without putting the observations themselves on a public ledger, which is how the authors keep confidentiality while still catching a silent edit.

Can Other File Formats Use the Same Method?

The authors say the approach is data-format-agnostic and name XBRL as a structured reporting format the same pipeline could cover. The paper’s later sections also list BBS+ selective disclosure and zero-knowledge proofs as future extensions, which would matter if a publisher wanted to prove a file is genuine without revealing every series inside it.

Disclaimer: This article is news reporting and analysis of a published research paper and an open-source prototype. It is informational only and is not investment advice, trading advice, or a recommendation to buy, sell, or hold XRP or any other digital asset. Readers should consult a qualified financial adviser or licensed investment professional before making any decision involving cryptoassets, and should read the BIS paper and code themselves before treating the prototype as operational guidance. Figures, test results, and project status reflect the cited sources as of the dates on those documents and may change if the authors, the BIS, or the ledger operators issue updates.