Summary

  • The Supreme Court of Mauritius restored an actionable route to AFRINIC board elections by confirming a receiver, imposing completion periods and later requiring the September poll to proceed in the presence of the Electoral Commissioner. That solved an authority and timing failure; it did not write every electoral rule.
  • The June 2025 design published committees, candidate steps, voting channels, identification requirements and a timetable. Yet it allowed electronic representatives, in-person representatives and proxies through several forms of corporate authority without publishing a sufficiently complete rule for resolving conflicting mandates or grading documentary defects.
  • After concerns about powers of attorney, the receiver annulled the June process while expressly saying the investigation had reached no final conclusions and that he could not report the extent of the irregularities. The gap between suspicion and institution-wide remedy exposed the missing evidence, materiality and review clauses in the plan.
  • The September replacement materially improved traceability: one newly designated voter per eligible member, direct executive authorisation, corporate-domain email, identity and biometric checks, a public register, online-only voting, three joint Election Trustees and court-ordered observation by the Electoral Commissioner.
  • The official record still does not provide a public observer report, a complete exception and rejection account, an independently explained technical tally assurance, or per-candidate totals in the result announcement reviewed here. Court presence improved confidence, but a durable election constitution requires a replayable evidence trail as well as a lawful supervisor.

The plan began where ordinary authority had ended

AFRINIC did not enter its 2025 election cycle as a normally governed membership company choosing a few directors in rotation. It entered after the board had lost quorum, director terms had ended, litigation had complicated corporate representation, and the Supreme Court of Mauritius had appointed a receiver to preserve the company and restore a board. The election was therefore expected to perform two jobs at once. It had to count votes, and it had to recreate the organ that would ordinarily have authorised and overseen the count.

That distinction explains both the strength and the weakness of court supervision. A court can identify a person entitled to act when the company's usual organs cannot. It can preserve assets, set a deadline, require reports and hear applications about the receiver's conduct. It can also add an external public officer to observe a contested poll. These are substantial interventions. Without them, AFRINIC could have remained caught between a constitution that expected a board and an institution that had no board capable of using it.

But an order to hold an election is not an election code. It does not automatically identify the complete electorate, reconcile every membership record, determine which officer may bind each corporate member, authenticate documents issued across dozens of jurisdictions, select a voting system, prescribe an evidentiary standard for objections or define the remedy for a defect discovered after voting. Those questions must be translated into rules by the temporary administrator and election bodies. They remain contestable even when the administrator's authority comes from a judge.

The 2023 order made this architecture visible. The Supreme Court directed that AFRINIC be preserved and that the value of the business be maintained. The Court of Civil Appeal's October 2024 judgment restored that order after rejecting an appeal brought without valid authority in AFRINIC's name. The appellate court treated reconstitution of the board as urgent and replaced the original completion period with two months from its judgment. A later order extended the practical deadline.

In February 2025, Gowtamsingh Dabee was appointed to replace the Official Receiver, and his April communique said the court had mandated him to reconstitute the board by election and then enable appointment of a chief executive.

The court had supplied a bridge across the authority gap. The quality of the crossing still depended on the design built on top of it.

Judicial supervision fixed the convening problem

The most important improvement was elementary: there was one identifiable office responsible for getting the election held. Before receivership, any proposed route back to a board could be challenged by asking who had power to call the meeting, instruct staff, approve expenditure, appoint committees or speak for AFRINIC. Once the court's order was restored, the receiver could point to a judicial mandate, the statutory powers associated with receivership and the company's constitutional documents.

That did not make the receiver equivalent to an elected board for every purpose. It made him the temporary centre of the restoration task. The April 2025 communique described the office as holding the ring, preserving the status quo of AFRINIC's assets and maintaining business value. It also claimed authority to organise, hold and complete the election. The language matters because it limits the purpose: the receiver's exceptional authority was justified by the need to return power to a functioning board, not by a permanent transfer of member sovereignty to an insolvency practitioner.

Court supervision also introduced a clock. The receiver announced 23 June 2025 as election day after an extension. When that process was annulled, he returned to court and obtained a new deadline of 30 September. Deadlines make delay observable. They require the temporary custodian to explain why a milestone was missed and prevent receivership from becoming an indefinite substitute for governance.

The judicial route supplied another discipline: contested decisions could be brought back before a body outside the election administration. ICANN's June 2025 application illustrates the boundary. The court said ICANN had no standing to enter the matter, yet it used its own authority to require a communique to members about Cloud Innovation's corporate registration and the constitution of the Nomination Committee. It refused the requested dismantling of the committee, finding that this would run against the election's purpose.

Whatever view one takes of the underlying dispute, the episode showed that external concern did not become a direction merely because it came from a global Internet institution. Coercive effect came from the Mauritian court.

This was a real gain. It separated persuasion from lawful command and gave members a forum in which the receiver's choices could be tested. What it did not do was make every choice self-proving.

The June design converted authority into visible machinery

The first 2025 plan was more detailed than a simple instruction to vote. The published election guidelines identified eight elected seats, a Nomination Committee, an Election Committee, a hybrid voting format, a third-party electronic voting provider and a sequence from nominations through results. Candidate eligibility was separated from voter eligibility. The Nomination Committee was assigned candidate criteria, interviews and the final slate; the Election Committee handled the election process; the receiver provided written guidance and support in the absence of directors.

The design also attempted to reduce conflicts. The receiver's April communique named a Nomination Committee chaired by an external lawyer and an Election Committee that included AFRINIC staff. The published materials said the Nomination Committee would not include candidates and would supervise the poll. Election administration and candidate selection were at least formally divided. The election guidelines stated that Nomination Committee members should be independent and neutral, while the Election Committee was responsible for practical conduct.

Voters were offered three channels: electronic voting through a designated representative, in-person voting through an authorised representative, or in-person voting by proxy. The plan required members to be in good standing, tied that status to completed membership formalities, current membership and paid fees, and used documentary checks for the people exercising corporate votes. The FAQ required a signed authority letter, a notarial certificate, a same-domain email and current identity evidence for an electronic representative.

Proxy appointments required an authority letter, notarial confirmation, a proxy form and identification, with physical documents to be produced on election day. In-person representatives also had to demonstrate authority and identity.

These were not trivial safeguards. They recognised that a vote cast for a corporate member had two authentication problems. The person had to be who they claimed to be, and that person had to possess authority from the member. An identity document answers the first question. A corporate resolution, director's letter or valid delegated authority answers the second. The June materials at least refused to treat access to an email address as sufficient proof of both.

The plan also published a voters-register communication and reminders close to election day. It warned that a member could use only one of the three voting methods. That rule addressed duplicate voting across channels. The institutional architecture was therefore visible enough to audit in broad outline.

Its vulnerability lay not in the absence of documents, but in the unarticulated law of proof between them.

Three voting channels created one authority problem in several forms

Hybrid access appears inclusive. A member unable to travel can vote electronically. A member represented in Mauritius can vote at the venue. A company can appoint a proxy when its ordinary representative is unavailable. In a geographically large service region, each option has a reasonable purpose.

Combining them in an emergency election, however, multiplied the number of boundaries at which authority could fail. The administrator had to know whether the member was legally eligible, whether fees and membership records were current, whether the person signing the authority instrument could bind the member, whether the representative's identity was genuine, whether an earlier designation had been revoked, whether a proxy instrument was valid under the law that governed the company, and whether the same member had attempted to vote through another channel.

The June FAQ recognised many of these questions but did not publish a complete priority rule for contradictory evidence. Suppose a registry contact designates an electronic voter while a director signs a later proxy. Suppose a power of attorney is valid for litigation or member meetings but ambiguous about AFRINIC elections. Suppose two corporate officers dispute each other's authority. Suppose a notary certifies a signature but not the signer's office. Suppose the company register has changed after AFRINIC's internal account was last updated.

A checklist can confirm that pieces of paper exist without deciding which legal proposition each piece proves.

The same problem applied to timing. Authority is not static. A director can resign, an officer can be removed, a mandate can be revoked and a proxy can be superseded. A credible plan therefore needs a cut-off, a hierarchy of authoritative sources, a revocation route, notice to all known corporate contacts and an exception process that freezes contradictory votes until the member's authority is resolved.

The June design did not publicly explain that hierarchy in enough detail. It specified inputs, but it did not provide a public decision table for conflicts among them. Nor did it publish the categories of rejection and the evidence needed to cure each category. Members could see how to submit documents; they could not fully predict how the administrator would decide when documents pointed in different directions.

This is the central lesson of the first poll. More documentation is not the same as a stronger evidentiary rule. A court-supervised process still needs to say what each document is competent to prove.

The committees were divided, but their dependencies remained joined

The 2025 materials used the familiar separation between NomCom and ElecCom. NomCom would encourage and assess candidates and supervise the poll. ElecCom would administer voter and election mechanics. The receiver would provide guidance, logistics and access to factual membership information. A third-party provider would operate the electronic system.

On paper, this distributes work. In practice, all four layers depended on information held by the institution being restored. Candidate-region checks relied on AFRINIC classifications. Voter eligibility relied on membership and billing records. Corporate contacts came from MyAFRINIC and related internal records. Staff supported the committees. The receiver appointed the bodies because no board existed. Independence of judgment did not create independence of data.

That dependency was unavoidable to some extent. An external committee cannot recreate twenty years of membership history from public material. The answer is not to exclude staff knowledge. It is to make the reliance testable. Every eligibility decision should identify the reference, its effective date, the person who reviewed it, any discrepancy and the authority for the final conclusion. Committee members should receive the same versioned dataset. Later changes should generate a visible exception record rather than silently altering the denominator.

The June public material did not show that full assurance layer. It named committees and described duties, but did not publish an independent reconciliation of the membership denominator into eligible members, registered voting representatives, issued electronic credentials, accepted proxies, in-person representatives, ballots cast and ballots counted. Without those joins, each body could perform its assigned step faithfully while the election as a whole remained impossible to replay.

Court appointment can reduce the risk that one faction convenes the poll without authority. It cannot remove the risk that every formally independent actor is looking at the same unverified institutional record.

June 23 exposed the missing law of electoral evidence

Voting took place on 23 June. Three days later, the receiver announced that the process was annulled after feedback and concerns about potential irregularities in voter documentation had been reported to relevant authorities. His more detailed notice of 15 July said suspicions had been raised particularly about powers of attorney used by some voters, that complaints had gone to the authorities and that police were investigating.

The notice contained an important qualification. The investigation had not reached final conclusions, and the receiver said he was unable formally to report the extent of the irregularities. That statement is responsible as far as it goes. It avoids presenting suspicion as a proven criminal fact and does not identify any person as having submitted a false document.

It also reveals the design gap. An election was annulled before the public record established how many instruments were disputed, which votes they affected, whether the concerns involved authenticity or scope, whether any disputed vote was counted, whether the possible number could alter a seat, whether the issue was isolated to a voting channel, or whether a narrower remedy was technically available. The receiver may have had confidential evidence that justified urgent action. The public materials reviewed here do not disclose it.

An emergency administrator sometimes has to act before a police inquiry is complete. Preserving legitimacy can require a pause when ballot custody or voter authority cannot be trusted. But the election code should specify the threshold in advance. It should distinguish a report from corroborated evidence, a defective form from an unauthorised vote, an affected ballot from an affected result, and a curable discrepancy from irreversible contamination.

The June plan said how to vote. It did not say enough about how to decide whether the vote remained certifiable after a challenge. That is why the annulment became an act of receiver discretion rather than the visible application of a pre-existing remedial schedule.

This article does not decide whether annulment was correct. The necessary evidence is not public, and a separate analysis is needed for the proportionality of any particular disputed proxy. The institutional conclusion is narrower: the plan should have made the path from allegation to preservation, investigation, exclusion, recount, partial rerun or full annulment legible before the allegation arrived.

An annulment announcement was not a reasoned election decision

The word "annul" carries more constitutional weight than the June communiques acknowledged. It does not merely postpone a meeting. It removes the effect of votes cast by members who may have complied with every published requirement. It resets candidate and campaign effort. It extends the temporary administrator's tenure. It can also change participation in the next poll, because some members will re-register and others will not.

A reasoned annulment decision should therefore answer a minimum set of questions even when sensitive evidence cannot be published. What power authorises the receiver or committee to annul? What category of defect occurred? What evidence was preserved? Which independent actor reviewed it? What is the possible affected-vote range? Could the defect have changed one seat, several seats or none? Why were exclusion, correction, recount or a partial rerun inadequate? What part of the decision can be challenged, before whom and by when?

The public notices answered only part of that set. They identified concerns about voter documentation, official complaints, a police investigation and the receiver's interest in transparency and fairness. They later recorded that the receiver reported the matter to the Supreme Court and obtained an extension. The court's willingness to extend the mandate established legal room for a replacement election. It did not publish the missing electoral findings.

That distinction protects the court as well as members. A court approving more time need not be understood as having decided that every allegation was true or that a full rerun was the only proportionate remedy. The receiver's own July notice said final conclusions were absent. Judicial permission to continue restoration and adjudication of the disputed facts are separate acts.

The institutional danger is retrospective certainty. Once a replacement election succeeds, officials may describe the first process as obviously compromised and the second as unquestionably legitimate. The contemporary record was more cautious. It contained suspicion serious enough to trigger annulment, yet limited public evidence public findings to explain the reach of that remedy. A sound history must preserve both facts.

The extension restored a deadline, not the lost evidentiary chain

After the annulment, the receiver returned to the Supreme Court. AFRINIC's 30 June update said he presented the underlying concerns and obtained an exceptional extension to 30 September. The 15 July notice added that the new process would reject both proxies and powers of attorney and would be conducted entirely online.

The extension performed a necessary continuity function. It prevented the failed June attempt from exhausting the receiver's authority and leaving AFRINIC without a route to a board. It also put a finite boundary around the second attempt. Court supervision was valuable here because it made the continuation lawful rather than self-authorised.

Yet a new deadline could not repair the first process's evidence record. The police inquiry remained unresolved in the receiver's public account. The full set of challenged documents, decisions and ballot effects was not published. The second poll therefore proceeded by redesign rather than by a publicly completed adjudication of the first.

Redesign can be sensible. Removing the disputed channel may produce a more reliable election faster than litigating every instrument. It should nevertheless be described accurately. The second plan was an institutional response to unresolved risk, not proof that every proxy or power of attorney in June had been invalid.

The September plan simplified corporate representation

The replacement guidelines made the most consequential change at the point where June had failed. Each eligible member had to designate one natural person to vote electronically. Every member had to designate again, even if it had participated in the annulled election. Proxies and power-of-attorney votes were not permitted. The designated voter had to be submitted directly by a company executive, use an email address from the member's domain and provide current identification.

This substantially reduced the number of authority paths. Instead of reconciling electronic representatives, in-person representatives and proxies, the administrator had one current designation per member and one voting channel. Re-registration prevented an uncertain June appointment from carrying over automatically. Confirmation messages to organisational contacts, described in the voter-onboarding material, created an opportunity for another officer to notice an unauthorised designation.

The revised structure was not perfect. "Executive" still needed interpretation across different corporate forms, and a company-domain email does not prove authority on its own. Some legitimate organisations may outsource email or use domains not reflected in older records. Corporate groups may have central officers acting for several entities. A strong exception procedure remained necessary.

Still, the design was easier to audit. It created a clean intended chain: eligible member, authorised executive, one designated voter, identity verification, platform activation, one submitted ballot and one included vote per open seat. Each transition could be logged. A discrepancy could be located at a particular boundary rather than spread across several voting methods.

The September rules also created three Election Trustees: the receiver, the ElecCom chair and the NomCom chair acting jointly. Automated results from the platform were to be made available to them for verification. Joint responsibility did not guarantee technical independence, but it reduced the chance that one office could silently open, close and certify the poll alone.

The improvement was procedural, not ideological. It did not settle every dispute about the legal meaning of AFRINIC membership. It made the act of exercising an accepted member's vote more direct and more observable.

Publication and biometrics improved traceability, with limits

The replacement process published a provisional voter register and then a final register. The official page said no discrepancies had been reported against the provisional version by the time of finalisation. It also said four designated voters were added after late confirmations from the nominating executives of eligible members, while no new registration applications were accepted after the deadline.

This is more transparent than a private roll revealed only after voting. A member could see whether its organisation and designated voter appeared. Competing corporate officers could raise an objection. Observers could compare the register with later participation statistics. The explanation of the four additions at least identified the category of exception: confirmation of applications already made, not acceptance of new applications.

The episode also shows why a change record matters. The public statement did not identify when each original application arrived, when confirmation was requested, who authorised the exception, which rule allowed it or how every known contact of the member was notified. Publishing those facts in privacy-preserving form would not require exposing identity documents. It would show that the same exception rule was available to similarly placed members.

Biometric onboarding added a further identity layer. AFRINIC's election statistics report 581 total voters, 548 completed biometric registrations, 33 who did not complete them and 484 votes cast. The final-register page said a designated voter could vote only after completing identity registration on the Voatz platform. These figures allow useful reconciliation: about 94.3 per cent of listed voters completed biometrics, about 88.3 per cent of those who completed biometrics cast a vote, and about 83.3 per cent of all listed voters cast one.

Those percentages describe participation and completion. They do not prove the validity of the underlying corporate designation or the secrecy and integrity of the tally. Facial matching can help establish that the person activating an account matches an identity credential. It cannot show that the company's signatory possessed authority to designate that person. Nor does a completed biometric check explain why an eligible member never reached the register.

The September design correctly treated identity as one control. Public accountability should resist turning it into a universal answer.

The Electoral Commissioner added observation at the decisive stage

On 5 September 2025, the Supreme Court ordered that the receiver proceed with the election in the presence of the Electoral Commissioner of Mauritius, who would oversee and supervise the 12 September process with the aim of a free and fair election in the best interests of AFRINIC and all its members. The receiver and Election Committee met the Commissioner the next day. The voting period was then set from noon on 10 September to 8 p.m. on 12 September.

This was the clearest addition of external electoral competence. The receiver had corporate-preservation authority. NomCom had candidate responsibilities. ElecCom administered the poll. A commercial platform handled electronic voting. The Electoral Commissioner brought experience from outside AFRINIC and was added by a specific court order after the first process had failed.

Presence can deter improvised changes. An observer can verify that the poll opens and closes at announced times, that the committee follows the published plan, that results are handled in an orderly manner and that complaints are escalated rather than suppressed. The court order also made it harder to describe supervision as a voluntary courtesy that the receiver could withdraw.

But the public record does not define the Commissioner's complete terms of reference. It does not say whether the Commissioner audited the voter-authority evidence, inspected platform configuration, observed cryptographic or administrative controls, reviewed excluded voters, witnessed the tally export, tested the result against cast-ballot records or merely supervised the visible closing process. Different forms of observation support different conclusions.

The Commissioner's presence should therefore be treated as an important safeguard, not as a substitute for an audit. A witness can attest only to propositions within the witness's mandate and access.

The missing observer report limits what supervision can prove

The official materials reviewed for this article announce the Commissioner's role, the consultation meeting, the revised voting period, aggregate election statistics and the winners. They do not include a linked public report from the Electoral Commissioner describing procedures observed, evidence inspected, exceptions encountered, limitations or a certification opinion.

That absence matters because "court-supervised" can otherwise become a conclusion rather than a description of authority. Readers may assume that every disputed layer was examined by the court or its observer. The actual order was narrower: the election was to proceed in the Commissioner's presence so that the process would be overseen and supervised. The court did not personally authenticate every voter or operate the platform.

A useful observer report need not disclose ballots or personal documents. It can state the observer's mandate, dates and access; the register version used; the number of credentials issued; the number activated; the number of votes cast; the opening and closing controls; the people present for each critical event; incidents and their disposition; reconciliation results; any departure from the guidelines; and the propositions the observer is not able to certify.

It should also distinguish observation from assurance. If the Commissioner watched procedural events but relied on the vendor for system integrity, the report should say so. If the Commissioner reviewed only aggregate results, that limit should be explicit. If no material incident occurred, recording that fact is more useful than silence.

The court's involvement made independent reporting possible. The election plan did not make publication of that report an explicit completion condition. That was a missed opportunity.

The aggregate figures illuminate turnout, not the whole electorate

AFRINIC's election statistics provide an unusually useful public funnel from listed voters to completed biometric checks to cast votes. They show where participation narrowed after registration. They also permit members to ask why 33 listed voters did not complete biometric registration and why another 64 people who completed it did not cast a vote.

No adverse inference should be drawn from those numbers alone. Some members may have chosen not to vote. A voter may have registered as a precaution but abstained. Technical or timing problems may have affected others. The figures do not classify reasons.

The more important denominator lies before the published 581. How many AFRINIC Resource Members existed for relevant purposes? How many were considered eligible under the court-referenced cut-off and good-standing rules? How many received designation invitations? How many submitted applications? How many were rejected, cured or left unresolved? The final register answers who reached a later stage. It does not, by itself, show the complete path from legal membership to electoral inclusion.

This article does not investigate individual exclusions; that requires member-specific evidence and a separate account. The design point is that a court-supervised restoration election needs aggregate reporting for each transition. A final roll can be accurate yet still conceal unequal notice, inconsistent cure opportunities or a failure to reach eligible organisations. Conversely, a low conversion rate can have innocent explanations. Only reason-coded counts allow a responsible conclusion.

The plan should have required a privacy-preserving reconciliation table after the poll. Categories might include ineligible under the stated cut-off, not in good standing, no response, incomplete executive designation, conflicting authority, identity failure, corrected before deadline, exception accepted, platform activated, ballot cast and ballot included. The public would then see the system's shape without seeing private files.

Court supervision is most valuable when it forces that denominator into view. Otherwise, observation begins only after the most consequential decisions about participation have already been made.

The result announcement did less than the guideline promised

Clause 9.1 of the replacement guidelines said the NomCom chair would announce the results, including the total number of votes for each candidate. The official AFRINIC announcement reviewed here lists the eight elected candidates by seat. The elected-candidates page does the same. Neither page provides per-candidate totals.

The election statistics page supplies aggregate participation: 484 votes cast. That is useful but different. Because each member could vote for each open seat, an aggregate count of voters cannot reconstruct the contest in any seat, the winning margin, the number of abstentions for a seat or whether any vote was excluded from a particular tally.

It is possible that totals were announced in a live closing ceremony or exist in another certified record. If so, the official result announcement should link to that durable record. Public accountability should not depend on locating a video, private email or temporary stream. The guideline created an expectation of numerical publication, and the permanent result should satisfy it.

Vote totals are not sufficient evidence of validity. They are, however, a basic consistency check. They allow members to compare participation, margins and any later correction. They also constrain rumours by fixing what the election authority certified at a particular time.

This is a small implementation gap compared with the June annulment, but it is revealing. Court supervision concentrated attention on getting a board elected before the deadline. A durable institution also needed a result record complete enough to support future review.

Evidence rules should have been written before the first designation

The two elections suggest a practical evidence schedule for any future emergency poll. First, the administrator should publish the legal electorate rule and versioned source date. Membership, good standing and voting eligibility must be separate fields, even when they often produce the same answer.

Second, the plan should identify who may designate a voter for each type of member and the hierarchy of proof. A current company register, constitutional office, board resolution and specific delegated instrument prove different things. A notarial certificate may authenticate a signature without establishing the signer's continuing corporate power. Every accepted instrument should be mapped to the proposition it supports.

Third, all known organisational contacts should receive notice of a proposed designation and any later change. Silence should not be treated as universal consent, but it creates a detection window. Conflicting instructions should move the member into a protected exception queue rather than producing competing credentials.

Fourth, the rule should define evidence states: received, incomplete, authenticity questioned, authority questioned, superseded, cured, accepted, rejected and under independent review. Each state needs a deadline and an appeal route. The public can receive aggregate counts; the member should receive reasons specific enough to correct an error.

Fifth, election data should be sealed in versions. The provisional register, final register, late corrections, credentials issued, activations, votes received and included tally must reconcile. Changes after the stated freeze require a named exception power, reasons and notice.

Sixth, the voting provider should supply assurance appropriate to the claims being made. The public does not need security-sensitive configuration, but it should know who controlled administrator accounts, how opening and closing were authorised, how duplicate credentials were prevented, how ballots were separated from voter identity, how totals were exported and how an independent reviewer confirmed the reconciliation.

These rules do not require a court to become an election technologist. They create a record the court can supervise.

A remedial schedule belongs in the plan, not the post-mortem

Every election code should assume that something will be challenged. The appropriate response depends on what is proved and where the defect sits. A late confirmation may require a reasoned exception. A failed identity check may require a cure period. A conflicting corporate mandate may require temporary exclusion and expedited review. A duplicate credential may require revocation before voting. A compromised ballot channel may require a rerun of that channel. A tally inconsistency may require recount and independent certification.

Full annulment should be available where the integrity of the result cannot be reconstructed. It should not be the undefined default for every serious allegation. The code should ask whether the defect is authentic, attributable, material, severable and curable. It should require preservation of the unaffected votes where technically and legally possible.

This is not a conclusion about the disputed June documents. The receiver said the investigation was incomplete, and the public record does not permit a reliable vote-impact calculation. It is a conclusion about institutional design. If the only clear choices are certification or destruction of the entire result, the plan has created an unnecessarily brittle election.

The review body also matters. NomCom supervised the poll, ElecCom administered it, the receiver held exceptional authority, the provider counted and the court supervised the receiver. A challenge should not return solely to the actor whose decision is challenged. The plan could designate an independent election reviewer, with urgent court directions available for questions of power and fairness.

Writing this schedule in advance protects the administrator from accusations of selecting a remedy to achieve a preferred outcome. It also protects members who complied with the rules from losing their votes without a reason tied to a published standard.

Courts should supervise boundaries rather than operate the ballot

AFRINIC's experience does not support the conclusion that judges should run regional registry elections. Courts are best placed to decide who has temporary legal authority, preserve assets and rights, enforce deadlines, compel disclosure and review whether the administrator stayed within the mandate. Election committees and qualified providers should conduct the operational steps. Members should choose directors.

The September order followed that separation. It did not replace ElecCom with the Electoral Commissioner or direct the judge to count votes. It added public supervision to a process administered by the receiver's election structure. That was a sensible response to a damaged trust environment.

The boundary fails if judicial language is used to shield operational decisions from scrutiny. A receiver cannot answer every question by saying the court authorised an election. The relevant inquiry remains specific: Did the court authorise this office? Did the guidelines authorise this act? Did the evidence satisfy the published test? Did the member receive notice and a review opportunity? Did the chosen remedy stay within the purpose of restoring governance?

The reverse boundary is equally important. Global technical institutions, governments and organised stakeholder groups can provide information and express legitimate concern. They should not be treated as possessing corporate command merely because AFRINIC's services have international effects. The June ICANN ruling demonstrated that the court could hear the concern while retaining authority over the remedy.

A strong court-supervised election therefore has three visible layers: legal authority supplied and reviewable under domestic law; electoral administration governed by published, technically competent rules; and member legitimacy produced by a verifiable vote. Collapsing the three makes supervision look stronger while actually obscuring responsibility.

The election record must survive the receiver

The September poll achieved the immediate institutional objective. AFRINIC announced winners for all eight elected board seats, and subsequent official communications described a board in place. The receiver later sought termination of the receivership while continuing to support transition pending formal discharge.

That success makes preservation more important, not less. The files should be sufficient for a future board, member or court to reconstruct the process without relying on the memory of the receiver, committee chairs, staff or vendor. At minimum, the archive should contain the orders, guidelines, committee appointments, conflicts, data versions, exception decisions, observer report, provider assurance, reconciliation, complaints, rulings and certified totals.

Handover should identify which decisions belong to the restored board and which remain fixed records of the election. The board should not be able to rewrite the evidence of its own election. Nor should the temporary administrator keep exclusive control of it after the office ends. Custody can be divided among AFRINIC's corporate records, an independent assurance holder and the court file, with personal and ballot data protected according to law.

The first task of restoration is often described as electing directors. The deeper task is returning an institution that can explain how those directors came to office. A result without a durable proof record leaves the next dispute dependent on the same emergency personalities the election was supposed to replace.

Court supervision restored procedure, not automatic legitimacy

The 2023-2025 record supports a balanced conclusion. Judicial intervention was not cosmetic. It answered who could preserve AFRINIC and convene a restoration election when corporate authority had collapsed. It imposed urgency, provided a forum for review, authorised a second attempt after the first failed and added the Electoral Commissioner to the decisive September stage. The replacement plan then reduced representation channels, strengthened direct designation and identity checks, published a voter register and completed an election before the extended deadline.

Those achievements should not be understated. A registry without a board acquired one through a process tied to domestic law rather than through unilateral selection by staff, a peer registry or an external Internet body.

Nor should they be overstated. The court did not resolve every membership boundary. The June plan did not contain a sufficiently public rule for conflicting corporate authority, evidence materiality and remedy. The annulment was announced before final investigative conclusions and without a published vote-impact account. The replacement made representation cleaner but left gaps in exception reporting, observer assurance, vendor-facing evidence and permanent result totals.

The right lesson is not that court supervision failed. It is that judicial authority and electoral integrity solve different parts of the same crisis. Courts can restore a lawful decision-maker and require fairness. Election rules must convert that mandate into equal treatment, reproducible evidence, constrained discretion and a record that members can test.

AFRINIC's plan succeeded at the first objective: it produced a route out of institutional vacancy. Its defects identify the next standard. A court-supervised registry election should be considered complete only when authority, electorate, representation, observation, tally and remedy can each be explained without asking the public to trust the title of the person who supervised it.

Sources and analytical limits

The starting authority is the Supreme Court of Mauritius receiver order of 12 September 2023, which directed preservation of AFRINIC and restoration of its board. The Court of Civil Appeal judgment in African Network Information Centre (AfriNIC) Ltd v Cloud Innovation Ltd and another, 2024 SCJ 473 is used for the restored receivership authority and revised election deadline. Neither source establishes that every later electoral decision was correct.

The first election design is taken from the receiver's 21 April 2025 election communique, the June election guidelines, the official election FAQ, the voters-register communique and the 18 June continuation notice. These documents establish announced requirements and procedures; they do not independently prove that each check was performed or that every underlying record was correct.

The court's June intervention is read from the Supreme Court minutes and ruling in the ICANN application and AFRINIC's final communique responding to that ruling. ICANN's official statements are evidence of ICANN's requests and description of events, not independent findings about disputed membership or election facts.

The annulment chronology comes from the receiver's 26 June communique, 30 June update and 15 July notice. The analysis preserves the receiver's express statement that investigations had not reached final conclusions. It does not identify any document or person as fraudulent and does not decide whether the full annulment was legally required.

The replacement design is based on the 2025 replacement election guidelines, candidate criteria, voter-onboarding announcement, final voter register, privacy notice and election statistics. They establish the published design and official aggregate figures, not an independent audit of the platform or member roll.

The added supervision is documented in the 5 September 2025 redacted order and AFRINIC's 6 September announcement. The outcome is taken from the official result announcement and elected-candidates page. The observation and publication gaps identified here are limited to the official materials reviewed and should be revised if AFRINIC, the Electoral Commissioner or the court publishes a fuller certification record.

The article does not determine current litigation, personal liability, the validity of a particular power of attorney, the entitlement of an individual member, the exclusion of any voter or the legal status of the elected board. Its proposed evidence and remedy rules are institutional recommendations derived from the documented difference between the two 2025 designs.