Summary

  • At 11:00 UTC on 13 September, LACNIC’s public 2026 Board-election page showed 12,609 eligible organisations across 30 countries. Its language-interface subtotals—3,661 Spanish, 708 English and 8,240 Portuguese—add up exactly.
  • The public calendar says the roll was published on 7 September at 18:05 UTC and that claims run from 8 September to 7 October. Yet the roll panel carries no visible observation time, revision number or change history.
  • This is not evidence that any entry is wrong. A claims process is supposed to correct a list. The missing control is a privacy-safe record connecting the first publication, each aggregate revision and the final roll used for voting.

The number 12,609 looks finished. The calendar beside it says otherwise.

LACNIC’s election system currently presents a searchable voters’ register for the three seats on its Board that begin new terms in 2027. Page one lists organisations by country, shows representatives only as initials followed by asterisks, and says it is displaying the first 50 of 12,609 records across 253 pages. The summary above the list gives 30 represented countries and separates the entries by the language configured in the interface.

That is useful disclosure. The three language counts reconcile to the total without residue. The list gives a member a practical way to check whether an organisation and its representative appear. Masking the representative’s surname limits casual exposure while leaving enough context for a member to recognise a record.

But this is not a static directory. LACNIC dates publication of the roll to 7 September at 18:05 UTC. Its claims period began the next day and remains open until 7 October. Article 25 of the bylaws explicitly permits claims, corrections and clarifications before voting and requires them to be settled. The institution has designed a process in which the list may legitimately change.

The public surface records the current state and the procedural window. It does not join the two. Beside 12,609 there is no “observed at” time, no roll version, no count of accepted or pending claims and no link to earlier totals. A member can see what the system says now, but cannot tell from the page whether it is the original publication state, a corrected state or the state that will later control voting credentials.

Precision without time is not provenance

A current counter answers one question: how many records does this view return now? An election denominator needs a second answer: which version of the roll governed a particular action?

The distinction matters even when every correction is proper. Suppose an organisation changes its designated representative, a duplicate is removed, an eligibility status is restored or a language preference is corrected. The live page can absorb each change and remain internally consistent. Its new total may be right. Yet a member who took a screenshot on publication day and a member who checks during the claims window would be examining different objects under the same title.

Nothing in the captured record shows that any such event has happened. Nor does the absence of a public revision log prove that LACNIC lacks internal audit records. The narrower finding is visible and reproducible: the public roll is open to correction, while its public presentation does not identify the revision being viewed.

The strongest defence is operational simplicity. Members need the latest roll, not a museum of every typo. Publishing named additions and removals could expose disputes, contact changes or eligibility details that should stay private. A long event log could also make an already large register harder to use.

That defence argues against publishing personal case files. It does not argue against aggregate provenance. A date, version and small change table can preserve the state transition without naming anyone.

One total, three different denominators

The roll count should also be kept separate from two later numbers. LACNIC gives members between one and eleven votes according to membership tier. The number of eligible organisations is therefore not the amount of eligible voting power. And neither number is the turnout: ballots returned and weighted votes cast are results of later actions.

Those distinctions are easy to lose when the first denominator has no durable state. A final results page might say how many organisations voted and how many weighted votes candidates received. To calculate participation or examine concentration, a reader also needs the frozen eligible-organisation count and the eligible weighted-vote total that applied when credentials were issued. A mutable pre-election webpage cannot supply that historical denominator unless its final version is identified and retained.

The displayed language split introduces another useful boundary. Spanish, English and Portuguese counts describe the interface setting attached to each listed record. They do not establish what language a representative speaks, where an organisation is incorporated, how many votes it controls or how it will vote. Their value here is mechanical: they provide a checksum for the current total. If the three subtotals cease to add up, the page would reveal an internal inconsistency. If they change together and still reconcile, only a revision record can explain the transition at an aggregate level.

A small receipt would be enough

LACNIC need not publish the contents of claims. A roll-version receipt could contain the version identifier, effective time, total eligible organisations, total eligible weighted votes and the three interface-language subtotals. For each superseding version it could add aggregate counts of records added, removed, corrected and still under review, plus the authority that approved the change.

The final receipt should identify the version used to issue voting credentials. Later results should refer back to that version rather than copying an unlabelled total. If a late correction is authorised, the new version should supersede the earlier one without erasing it.

This would complement, not replace, the searchable register. The live list remains the service members use; the receipt becomes the evidence that lets them compare what was published, what changed and what finally governed the ballot.

There is no basis here to say that 12,609 is wrong, that an eligible member is missing or that anyone has been improperly included. Indeed, the visible arithmetic is clean. The problem is more ordinary and more repairable: a precise electoral number is travelling through a correction process without carrying its own clock.

Sources