First verification pass on the register index
Forty-seven registers checked by hand. Denmark amended its law so the register records why an approval ended instead of deleting the person. The European Union legislated most of this for one trade in 2006, Italy did it for every regulated profession in 2012, Poland deleted its register in 2014, Estonia has every component of a signed register except the signed file, and three jurisdictions keep the register that counts in a room.
The index now carries forty-seven registers across thirty-three jurisdictions and forty-one trades, each checked against the authority’s own published pages on 9 November 2026, with what was found written out on the row’s page. Health, law, finance, construction, electrical work, gas, security, architecture, migration, property. Several findings were worth the pass on their own, and two of them changed what this lab thinks it is asking for.
France was wrong in the lab’s own notes. The working assumption had been that no national public register of estate agents existed. It does: a 2015 decree created a national file of carte professionnelle holders and put CCI France in charge of it, after the card moved from the prefectures to the chambers of commerce. Public, consultable, no export. The grade changed and so did the argument.
Queensland publishes its builders as a CSV and refuses to publish the older versions, stating that it does so to avoid putting out data that is no longer correct. The reasoning is sound and the consequence is severe: you can check a licence today and you cannot establish what the register said on the day the money moved, which is the only question anybody asks after a fraud.
Ahpra prices the machine-readable route at four thousand dollars to connect and a thousand a year to keep, and that paid feed refreshes every 24 hours while the free public page refreshes about every fifteen minutes. The programmatic route is slower than the human one and costs money. That is not an authority being obstructive; it is what happens when data access is built as a product for employers rather than as a public artefact.
Texas is the counter-example that makes the whole argument arguable. The Department of Licensing and Regulation publishes a CSV per programme, updated daily, plus an all-licences file covering everything it regulates, with no key and no account. Tier A of RVW-5 is not a proposal there; it is a description of what one authority already does. Nobody had to build a protocol for it to happen.
Four more are worth stating plainly. The UK sells its architects: anyone may search the register free, and a copy of the whole thing is a paid PDF behind a link that expires, with the website and the purchase download listed as two separate items in the board’s own publication scheme. The UK security regulator will not let you search without the licence number unless you supply the person’s date of birth, so checking a credential means holding personal data about them. Australia’s migration authority warns applicants, in its own user guide, to be careful of people pretending to be registered agents, which is this lab’s problem statement written by a government office. And Germany registers more of its tradespeople than almost anywhere while letting the public check almost none of them: compulsory chamber membership, no national lookup, enquiries by telephone, and the enquirer expected to show a legitimate interest.
Then Singapore, which is the one that should be uncomfortable reading for everybody else in the table. The Council for Estate Agencies publishes every valid salesperson as a CSV of around thirty-eight thousand rows, refreshed three times a day, through a documented query API, under a licence that says free forever for personal or commercial use. It publishes the suspended and the revoked as separate lists. It publishes a rolling three-year record of enforcement actions. It publishes the transactions each agent actually completed. Everything this lab spends specifications arguing for, minus the signature, one regulator simply decided to do.
And Dubai, which has gone furthest on the thing itself. Every broker carries a registration number, every advert must carry a permit number issued by the regulator, and the advert can be scanned and checked against the registry on the spot. A listing is bound to a licensed person and the binding is enforced with fines, suspension and permanent cancellation. It works, and it is not portable: no open file, the machine route behind a gateway, and the proof living inside one authority’s app rather than in an artefact anyone can verify for themselves. That gap is the whole argument for doing this as a protocol instead of as a product, and it is better evidence than anything the lab could have written.
Two more worth recording. Ireland keeps the statutory register of practising solicitors inspectable at the Society’s premises in Dublin during office hours; what is online is a directory whose own disclaimer says it should not be regarded as a comprehensive list. The Society has had to post a notice that an AI search tool is producing inaccurate information about disciplinary findings, which is what happens when the authoritative record is the hardest of the three to reach. The Netherlands, by contrast, documents its health register’s machine interface down to the WSDL and publishes the coupling under CC0, and still offers no file, so there is nothing to hash and nothing to sign.
British Columbia deserves the last word, because its regulator wrote the best sentence in the index about its own public licensee search: this is an unofficial search. It shows the current licence period and not the licensing history, and anyone who needs an official answer is told to telephone. The same authority runs a consumer campaign telling people to search before they sign, pointing at that search, while agreeing to give a regional realtor association on-demand access to its licensing data in exchange for transaction data. The industry body gets a feed and the public gets a disclaimer. Scotland, by contrast, has most of the shape already: a card, a five-digit number, and a status check that tells anyone whether that number is currently entitled to practise, and says not entitled to practise when it is not. No file, no export, and an answer you cannot keep.
Three findings from the later passes deserve to be read together, because between them they answer most of the objections this lab gets. Italy legislated the hard part in 2012. INI-PEC is a national index, established by statute, holding the certified electronic address of every company in the business register and every professional enrolled in an order, fed by the orders themselves, free and searchable with no credential, and the Digital Administration Code states that consultation of the digital domicile registers is freely permitted, not subject to authentication, and that the lists are produced in open format. Everyone who says no government would ever mandate that every licensed professional publish a verified digital address in one national index should know that one already did, and that the address carries legal weight when something is sent to it. It binds an inbox rather than a key, and that is the only real difference.
Estonia answers the technical objection and sharpens the ask. It has a register of professional certificates refreshed nightly, identities validated against the population register, a citizen view of your own certificates including expired ones, and every message between those systems digitally signed and timestamped with logs admissible in court. Citizens receive signed files from public bodies as a matter of routine and verify them on a laptop. Every component of grade A exists in that country and the grade is still B, because the state signs the channel between institutions and signs documents issued to individuals, and publishes no signed file of who holds what. Sign the artefact, not the channel. Estonia is the proof that those are not the same thing.
Poland is the controlled experiment nobody would have been allowed to run. On 1 January 2014 the central registers of estate agents and property managers lost legal force and were closed, pending applications were discontinued, and the licence, the clean record, the education requirement and the placement all went with them, along with professional liability. Trade federations launched competing voluntary registers to fill the hole. Twelve years on the state is moving back the other way, restoring a central register of property managers with entry required in order to practise, on the stated ground that deregulation in property produced negative effects. When somebody asks what a public credential record is actually worth, that is the citation.
A late addition is worth one line because it states the opposing case better than any critic has. Norway publishes its centrally approved construction firms through a free REST API with no key and no registration, which is as open as this index gets, and asks users not to store the data permanently but to look it up live instead. That request is entirely reasonable: it keeps every answer fresh and it stops stale copies circulating. It is also exactly what a durable proof cannot do. No copy means no provenance, and no way to show, a year into a dispute, what the register said on the day the contract was signed. Freshness and provenance pull in opposite directions, and a signed dated file is the only thing that gives you both.
Three later rows say more about the specifications than anything the lab has written. Denmark used to delete an estate agent from its public register when their approval ended. The law was amended so that the approval is instead recorded as deposited, lapsed, withdrawn or revoked. Somebody worked out that erasing a person destroys the only record that answers the question a disputed transaction actually asks, which is not whether they are approved today but what the register said then, and why it stopped. That is RVW-6, passed into law, by a jurisdiction that had never heard of this lab.
Seoul, facing a wave of lease-deposit fraud, has said it intends to publish detailed business information about practising agents so that risky ones can be identified, and to do it with the agents’ consent. Consent-based enrichment of a public register, for fraud prevention: that is tier C of RVW-5, reached independently, by a city government with a problem rather than a protocol.
And Ireland supplies the evidence that impersonation is a real offence rather than a hypothetical. Its utilities regulator has prosecuted people for carrying out gas work while falsely claiming to be registered, including one whose claim was disproved by a records check and, most pointedly, a former registered installer who kept claiming registration after his had been terminated. The credential was real, then it was not, and the only thing between the public and the lie was somebody thinking to check. An ID card with a photograph answers that in a doorway. It answers nothing over email.
What this pass does not have yet: archived copies of the evidence with their digests published beside each row. Until that exists a row can be disputed but not checked against a frozen copy, and that is the next piece of work. Corrections are wanted, in particular from anyone who works inside one of these forty-seven.