Summary

  • AFRINIC’s August 2026 proposed Constitution does not create the Article 11.4 emergency-policy power. The current Bylaws already contain materially the same provision, and the proposal’s own rationale says Article 11 receives no substantive change beyond editorial and consequential amendments.
  • Article 11.4 lets the Board decide for itself when an Internet-number-resource policy is “necessary and urgent” and move outside the ordinary Article 11.3 route. The clause supplies no trigger taxonomy, evidence burden, contemporaneous reasons, fixed sunset or immediate independent review, even though registry action can quickly affect the shared records and services on which running networks rely.
  • A narrow, reversible emergency mechanism can be justified when uniqueness, registry integrity, security integrity or continuity faces a demonstrable threat. It cannot convert AFRINIC, a private bookkeeper and coordinator, into a sovereign, owner, police force or punishment authority; participation and Board procedure remain evidence of institutional activity, not a source of public power.

The reform that preserved the exception

The most consequential word in AFRINIC’s August 2026 proposed Constitution may be “notwithstanding.” Article 11.3 sets out the ordinary route for Internet-number-resource policy: proposals are discussed and agreed at an open Public Policy Meeting through the Policy Development Process defined by the Regional Internet Community, and are then ratified by the Board. Article 11.4 begins by stepping around that route. Where the Board considers a resource policy necessary and urgent, having regard to what the text calls proper and responsible resource usage, the Board may adopt it without first completing the ordinary process.

That is a substantial decision surface. Yet the August draft does not newly open it. Article 11.4 already appears in AFRINIC’s current 2020 Bylaws in materially the same form. Earlier text carried a mistaken cross-reference to Article 11.2; a 2020 special resolution corrected that reference to Article 11.3, and the August draft keeps the correction. The proposed Constitution’s rationale removes any residual doubt about the drafter’s account of the change: it says that Article 11 has no substantive amendments apart from editorial and consequential ones.

The accurate description, then, is not that the August draft invents an emergency power. It retains one. A constitutional review capable of reconsidering the organisation’s governing machinery has carried forward the clause that lets the Board move first under a standard the Board itself applies. The reform question is therefore sharper than a debate about novelty. It is why a provision with such immediate operational reach survived a rewrite without acquiring objective triggers, an evidence rule, a reason-giving duty, a calendar-based expiry or an independent route for urgent challenge.

The draft remained a proposal at the date of this article. AFRINIC describes its Bylaws Review Committee as advisory and its recommendations as nonbinding, with final amendments to be presented for approval at a Members’ General Meeting. The published consultation schedule contemplated further comments, a final report and a later members’ meeting; it described dates as indicative. No checked public record by 10 August 2026 established that members had adopted this proposed Constitution or that it had taken effect.

It would be equally wrong to write as though an emergency policy had already been made under the proposal: no completed invocation under this draft, implementing manual or measured outcome was located in the checked public materials.

Those uncertainties do not make the text inconsequential. A proposed constitution tells members what powers its authors are prepared to preserve. Article 11.4 is especially revealing because the draft labels its continuity as nonsubstantive housekeeping. A power can be old and still be insufficiently bounded. Indeed, carrying it through a formal review may make the omission of safeguards more important, because the moment for deliberate correction is visible and the underlying risk is no longer an accidental remnant hidden in inherited language.

What the ordinary route protects

The ordinary Policy Development Process is not valuable because every meeting is wise, every participant is representative or every consensus call is beyond dispute. Its value is more practical. It separates proposal from adoption in time. It exposes assertions to people with different network conditions and commercial incentives. It creates a record in which an operator can ask what problem is being solved, what evidence supports the claimed problem, which systems would change and what collateral effects have been ignored.

Open discussion can reveal when a general rule has been designed around an exceptional dispute, or when a policy that seems harmless at the registry desk would impose expensive changes on live networks.

Participation does not confer sovereignty. A mailing list, meeting room or broad class of interested people is not a legislature merely because it is open, and neither attendance nor rough agreement gives a private registry ownership of addresses, networks or commercial decisions. Participation is evidence. It can show operational acceptance, uncover errors, supply competing facts and test whether a coordination rule remains thin enough to serve a shared technical function. Those are important benefits without the mythology that “the community” can manufacture authority by gathering.

Article 11.4 displaces the timing of that evidence-testing. Under the clause, adoption can precede the ordinary discussion and agreement process when the Board considers the policy necessary and urgent. The shift matters because adoption is not a neutral draft circulation. Staff may receive an implementation direction; members may prepare to comply; systems may be configured; counterparties may treat the registry’s position as an operational fact. Even if later deliberation is possible, the decision arrives first, and the burden of challenging it falls on those already reacting to its effects.

Draft Article 11.5(a) provides a later endorsement checkpoint at the next Public Policy Meeting. Its existence matters, but it does not answer the front-end questions inside Article 11.4: what event qualifies, what the Board must prove, what must be disclosed at adoption, what may be changed before review, and how an affected operator can obtain a prompt independent assessment. The existence of a later meeting is not a substitute for defining the emergency at the moment exceptional power is used.

This is the structural bargain hidden inside the word “notwithstanding.” The organisation trades prior, distributed testing for speed. A defensible emergency clause would therefore compensate for the lost safeguards with unusually precise triggers and unusually strong real-time accountability. Article 11.4 does the opposite. Its triggering terms are broad, and its emergency-specific procedure is thin.

The Board grades its own emergency

The clause turns on where the Board “considers” a policy necessary and urgent. That language makes the Board the textual judge of both predicates. Article 11.4 names no separate technical fact finder, no independent reviewer and no affected-party hearing before adoption. It does not say that necessity or urgency must be established to a defined evidentiary standard. Nor does it require the Board to identify the harm that would occur if the ordinary process ran its course.

The gap begins with the absence of a trigger taxonomy. The draft does not confine emergency use to a duplicate allocation, a verified compromise of registry systems, a cryptographic integrity failure, a breakdown threatening unique coordination or another functionally testable incident. It does not distinguish imminent operational harm from an important but ordinary policy preference. “Proper and responsible usage” is not converted into an observable test. Without such conversion, the phrase can carry whatever normative judgment the decision-maker places inside it.

Necessity is also left unstructured. The clause does not require proof that the proposed action is the least disruptive means of preserving a registry function. It does not demand consideration of a narrower technical containment measure, a temporary preservation of the last verified state or a parallel consultation that could run without delaying a minimal intervention. A Board could sincerely believe that a broad policy is necessary while never having to show why a limited and reversible measure would be inadequate.

Urgency receives no more discipline. There is no required account of the time available, the point at which harm becomes irreversible, or the reason ordinary Article 11.3 discussion would be too slow. The difference between “important now” and “cannot safely await the ordinary route” is the heart of any emergency test. Article 11.4 leaves that difference to institutional judgment rather than making it a public evidentiary burden.

The text likewise does not require a contemporaneous written decision containing the trigger event, supporting material, alternative options, expected effects, implementation time and review deadline. It does not require publication of dissent or transaction-specific conflicts when the policy is adopted. It does not require an impact assessment or immediate notice to operators whose registry position or services may change. It contains no emergency-specific route to an independent reviewer able to examine the evidence while intervention can still be limited or reversed.

That arrangement creates a self-validating loop. The same institution identifies the emergency, judges the evidence, selects the scope and authorises implementation. If the only operative test is that the Board considers the measure necessary and urgent, the conclusion risks becoming its own proof. A declaration of urgency can then be used to justify the exceptional route, while the use of the exceptional route gives the declaration practical force.

The problem is not that directors are incapable of acting honestly. It is that good institutional design does not depend on confidence in the particular people occupying a role. Honest decision-makers can be mistaken, can see evidence through the organisation’s incentives and can underestimate the costs borne by parties outside the meeting. Emergency powers should be written for the moment when facts are contested, interests diverge and trust is lowest. A clause that works only when everyone trusts the Board is weakest exactly when the emergency label becomes most consequential.

General corporate controls are not emergency controls

The draft is not devoid of Board procedure. Its general rules define the Board as directors acting together with the required quorum. Article 19.6 sets a majority quorum with a floor of five directors. Article 19.8 uses a majority of eligible votes cast. Article 19.10 requires accurate minutes and provides for publication of approved minutes, subject to redaction. Article 19.13 requires an organisation-wide conflict-of-interest policy. Matters for decision are also subject to general notice requirements.

Those provisions are relevant. A quorate meeting is better than an informal instruction; a recorded vote is better than an untraceable decision; a conflict framework is better than silence; published minutes can let members reconstruct part of the institutional history. But none of those general controls answers the questions created by Article 11.4.

An ordinary majority is not a heightened threshold for displacing the ordinary policy route. A general conflict policy is not a clause-specific requirement to disclose which director has an interest in the emergency policy, who recused, how the remaining quorum was calculated or whether an independent person tested the conflict. Minutes prepared and approved later are not the same as a contemporaneous public case issued when the exceptional policy takes effect. Minutes may record that a resolution passed without publishing the evidence needed to test necessity, urgency, scope and proportionality.

Timing is especially important. An operator deciding whether a registry change threatens routing-adjacent operations cannot wait for the normal cycle of minute approval to learn why the action occurred. A lender, customer or technical partner assessing continuity risk needs to know the immediate scope and rollback conditions. An independent reviewer needs the record before operational reliance and institutional commitment harden around the decision. Retrospective transparency can inform accountability, but it cannot fully restore a lost chance to prevent an overbroad first move.

The proposal also contains a tension in its own allocation of policy functions. Draft Article 15.3(a) says directors may determine address-allocation guidelines in line with the member-driven Policy Development Process. Article 11.4, however, authorises a resource-policy route notwithstanding the ordinary Article 11.3 process. The draft does not provide an emergency-specific rule for reconciling the commitment to a member-driven process with the power to adopt first.

A later checkpoint exists, but the conflict at the moment of adoption remains: the extraordinary path borrows legitimacy from a policy system whose normal sequencing it displaces.

Corporate formalities cannot cure that tension by themselves. Quorum and majority voting tell us that a decision was institutionally made. They do not establish that the asserted emergency was real, that the action was technically necessary, that its scope was minimal, or that the organisation possessed authority to impose every consequence written into the policy. Procedure authenticates an internal act; it does not expand the lawful or technical perimeter of the actor.

The emergency clock has no fixed face

Article 11.4 states no number of days for which an emergency policy may operate. It has no automatic expiry and no event-based sunset. The later checkpoint in Article 11.5(a) is tied to the next Public Policy Meeting, while Article 11.2 requires at least one such meeting each year. That establishes a minimum meeting frequency, not a fixed lifespan for an emergency policy.

The distinction is more than drafting neatness. A deadline expressed as a date or short period runs independently of the institution’s convenience. A deadline tied to “the next meeting” depends on the calendar. The interval may vary with when the emergency decision occurs, when a meeting is scheduled and whether the published timetable changes. The checked text does not justify converting that arrangement into a claim that every emergency measure lasts 30, 60, 90 or 365 days. No such fixed maximum appears in Article 11.4.

When the decision-maker can both declare the trigger and operate within a schedule-dependent duration, it gains two forms of control: control over entry into the exception and practical influence over how long the exceptional interval persists. Even without improper intent, that design creates institutional drift. Temporary arrangements acquire implementation teams, internal owners, member expectations and counterparties that rely on the new state. Reversal becomes more costly with every dependency added around the policy.

A genuine sunset should therefore be calendar-independent and short enough to force a fresh evidentiary decision. Expiry should occur automatically unless an independent process authorises a narrowly defined continuation. The clock should begin at adoption, and the public record should state it. Later community examination can coexist with that mechanism, but it should not be the only temporal boundary.

The absence of a fixed sunset also changes the incentives at the start. If directors know that an intervention will lapse quickly unless defended again, they have reason to choose a minimal, reversible measure and prepare evidence from the outset. If the measure can persist until a schedule-defined event, the incentive leans toward adopting a complete policy and allowing implementation to create its own momentum. Time limits are not merely safeguards after a decision; they shape the ambition of the decision before it is made.

The strongest case for moving fast

There is a serious argument for an emergency route. A shared registry cannot always wait through a long public process if unique state is actively compromised, two records collide, a security failure corrupts trusted information or a time-critical coordination fault threatens live-network continuity. The ordinary route is designed for deliberation, not incident response. In such a circumstance, delay can multiply reliance on an incorrect state and make a later correction more disruptive.

A small group of accountable directors can also convene faster than a wide community. Staff with direct access to registry systems may possess evidence that demands immediate containment. Public disclosure might need careful staging if it would expose a live security weakness. A minimal intervention that isolates corrupted state, preserves logs and prevents further conflicting changes can protect operators while an open process catches up. Refusing any possibility of rapid action would mistake slowness for legitimacy.

But the strongest argument for speed proves only the need for a narrow continuity mechanism. It does not prove that “proper and responsible usage” is a sufficient trigger for policy-making. It does not prove that the Board should be sole author, fact finder and judge. It does not prove that reasons can wait for ordinary minutes, that affected operators need no immediate review or that the registry may use an operational incident to regulate commercial morality.

The technical case for emergency action contains its own limits. If the threat is a collision in unique records, the response should isolate the collision and preserve the last verified state. If the threat is a registry compromise, the response should protect integrity, evidence and continuity while disabling only what the compromise requires. If the threat is a security-integrity fault around publication, the response should be narrow, reversible and auditable. The farther a policy moves from a measurable failure of uniqueness, integrity, proof of control, interoperability or continuity, the weaker its claim to emergency treatment becomes.

The institution should bear the burden of showing why the ordinary process cannot safely run. That record can be concise when time is short, but it cannot be empty. It should identify the invariant at risk, the evidence that the risk is real, the expected harm during ordinary deliberation, the narrowest available containment and the time at which independent review will occur. A security-sensitive annex may sometimes be necessary, but confidentiality should protect specific exploit details, not erase the existence of a reasoned decision.

Speed and accountability are not opposites. Review can run in parallel with a temporary technical action. Notice can follow immediately after containment. An independent person can have power to stay an overbroad measure without operating the registry. A short sunset can force reconsideration without requiring the whole community to solve an incident in real time. Reversibility can preserve action under uncertainty while keeping mistakes from becoming permanent facts.

AFRINIC’s power stops at the ledger

The deepest boundary is institutional, not procedural. AFRINIC is a private bookkeeper and coordinator. Its legitimate common function is to maintain thin shared state: unique registration, accurate records, proof and contact information, security integrity and operational continuity. It is not a sovereign, legislature, regulator, police force, prosecutor, court, punishment authority or owner of the networks and number-resource capital recorded in its systems.

That boundary does not disappear in an emergency. A corporate constitution can allocate functions within a company, subject to applicable law and contract. It cannot create public jurisdiction by declaring that directors may act urgently. A Board vote cannot turn registration into title over an operator’s infrastructure. A broad community meeting cannot confer police power on a registry. Participation supplies evidence of technical acceptance and practical effect; it does not manufacture sovereignty.

The distinction matters because the registry’s records are operationally influential. Other actors may rely on WHOIS or RDAP data, registration status, contact records, resource certificates or related services. A registry instruction can therefore feel coercive even when it is not law. The dependence of operators on shared records creates leverage, but leverage is not authority. It is a reason to constrain the bookkeeper more tightly, not a reason to treat the bookkeeper’s practical power as self-justifying.

An emergency policy may preserve accurate registry state. It may temporarily halt a conflicting clerical change, secure an audit trail or maintain continuity while facts are tested. It may not punish an operator for alleged misconduct, confiscate operational value, decide universal ownership, impose unrelated service withdrawal, or convert an internal view of responsible use into a territorial command. Questions of legal rights and enforceable sanctions belong to competent legal institutions with jurisdiction and due process.

This is also why “responsible usage” is a dangerous open phrase in an emergency clause. Within a narrow registry function, responsibility can mean maintaining accurate evidence, avoiding duplicate state and protecting the integrity of a shared technical system. Outside that function, the phrase can become a container for moral, political or commercial judgments that the private registry has no mandate to enforce. The constitutional text should draw that line, not leave the Board to discover it after declaring urgency.

Running code is the primary continuity interest. Administrative changes should protect networks that are operating, customers who depend on them and technical systems that have built legitimate reliance on the last verified state. Where facts are uncertain, the default should be preservation plus auditable containment, not destruction of operational value. A reversible ledger action can be corrected; a measure that interrupts services, damages routing-adjacent trust or collapses financing around an operator may create losses that a later institutional acknowledgment cannot reconstruct.

How an internal rule reaches an operator

Article 11.4 concerns policies for managing Internet number resources, so its practical reach can extend across the points where AFRINIC’s systems meet operator activity. A fast policy could change intake, eligibility, allocation processing, transfer recognition, contact duties, registry status, WHOIS or RDAP publication, security-integrity handling or service treatment. The exact effect would depend on the policy; no particular policy or implementation is established here. The point is that the constitutional route can accelerate a change across surfaces on which operators rely.

That acceleration creates hold-up risk. An operator may have contracts, customers, financing and technical plans built around a stable registry position. If an emergency policy changes how that position is recorded or serviced before ordinary discussion, the operator must respond immediately even while disputing the policy’s basis. Counterparties may react to the registry’s published state before any independent reviewer examines the evidence. Compliance and contingency costs rise with uncertainty, especially when the trigger and duration are undefined.

No reliable numerical estimate can be attached to that risk from the checked materials. There is no verified count of Article 11.4 cases under the proposed draft, no distribution of emergency durations, no measured implementation bill, no affected-address total, no financing discount and no outage-loss figure. Inventing a dramatic number would weaken the analysis. The direction of the mechanism is enough: broader triggers, less evidence, slower review and an uncertain end date increase the chance that an internal decision becomes costly to resist and difficult to unwind.

Reversibility is therefore an economic control as much as a technical one. A temporary flag with preserved prior state is easier to correct than deletion. A limited pause on conflicting clerical changes is easier to unwind than a policy that restructures multiple relationships. Immediate notice lets an operator protect customers and preserve evidence. A published rollback path reduces uncertainty for counterparties. Independent review can stop implementation before reliance spreads.

The Board should also distinguish a registry event from a legal dispute. A court order within jurisdiction may require action, and an actual integrity failure may demand technical containment. But an allegation, commercial conflict or contested claim does not become a registry emergency merely because the registry can make one side’s position operationally painful. The ledger should reflect verified facts and lawful orders; it should not be used as a substitute courtroom.

The same restraint protects AFRINIC. An institution that confines itself to documented bookkeeping can explain its action in technical terms and reverse it when facts change. An institution that uses urgency to choose winners, discipline conduct or claim ownership assumes risks it cannot competently resolve. Thin coordination is defensible because everyone needs unique and accurate shared state. Thick governance invites every contested political and commercial issue into the registry.

A proposed clause in a contested governance setting

The institutional setting increases the need for precise language. NRS’s formal position is that the individuals presently presented as AFRINIC’s Board have not established lawful authority merely through an election announcement or company-registry filing while litigation and the discharge of the Receiver remained unresolved. That is an important accountability position, not a final judicial determination supplied by the available record.

AFRINIC’s official court index proves that proceedings are listed and records how the organisation describes their status; it does not by itself prove the merits of a judgment that has not been established here.

Where current authority is material, the body is therefore properly described as the purported Board. That wording does not decide the litigation. It prevents the organisation’s own labels and filings from being treated as self-proving. Official AFRINIC materials are evidence of words, acts, documents and internal rules. They can establish what a draft says, what meeting minutes record and what process the organisation announced. They cannot, by their own publication, finally establish legitimacy, representation, stewardship or jurisdiction.

NRS urged members not to treat participation or votes as recognition of the purported Board and sought records supporting authority. It has also warned that a Bylaws rewrite amid unresolved disputes could consolidate institutional power. Those positions deserve accurate attribution and scrutiny, but they do not prove that the proposed Constitution has been adopted, that Article 11.4 has been invoked or that a court has reached a final result.

NRS acts as advocate, researcher, convener and representative of members who expressly authorise it; it does not operate AFRINIC’s registry, policy process, WHOIS or RDAP, security systems, appeals, elections, custody or continuity functions.

The answer to disputed authority is not to pretend that corporate governance can be ignored. Nor is it to assume that more participation cures every defect. The answer is to demand evidence at each layer: evidence that the directors are lawfully in office and quorate, evidence that a constitutional text is effective, evidence that an Article 11.4 trigger exists, evidence that the response is necessary and narrow, and evidence that affected parties have access to independent review. None of those propositions should borrow proof from the others.

This separation is essential during an emergency. Operational continuity may require staff to preserve systems and verified data while authority questions are heard. Preservation is not endorsement. Keeping accurate shared state does not concede that a purported Board may enlarge its power. Likewise, challenging authority does not justify damaging the registry services on which networks rely. Continuity and accountability can be pursued together if emergency action remains clerical, minimal, logged and reversible.

The safeguards the text should have named

A credible Article 11.4 could begin with an exhaustive or functionally precise list of triggers. These might include a demonstrable threat to unique registry state, an active compromise of registry integrity, a failure of proof-of-control systems or an immediate continuity fault. The list should be tied to minimum coordinating functions, not to a general view that an operator’s use is undesirable. If a claimed objective can await the ordinary Policy Development Process without demonstrable harm to uniqueness, integrity or continuity, it is not an Article 11.4 emergency.

The Board should then carry an explicit burden. A decision record should identify the event, evidence, threatened invariant, expected harm during ordinary process, alternatives considered and reasons a narrower action would fail. It should state the legal and contractual basis, the precise registry surfaces affected, the implementation time, the responsible staff, the rollback method and the automatic expiry. It should disclose director conflicts and recusals, record dissent, and explain any redactions. Sensitive technical details can be protected while the public reasoning remains available.

The action itself should be minimal and reversible. The first move should preserve the last verified state, isolate conflict, secure evidence and maintain running-code continuity. It should avoid deleting records, extinguishing operational value, withdrawing unrelated services or making irreversible allocations of advantage while facts remain disputed. If a broad policy is eventually warranted, the ordinary process can consider it after the immediate threat has been contained.

Notice should be prompt and specific. Affected operators need to know what changed, why, which evidence may be challenged and where to seek review. An independent reviewer should be available immediately, with power to examine authority, evidence, proportionality and continuity and to stay any part of the measure that exceeds minimal containment. Independence cannot mean another body whose mandate or information depends entirely on the same decision-makers.

Finally, a fixed expiry should operate independently of meeting schedules. Any continuation should require a new, published evidentiary decision rather than inertia. The later endorsement checkpoint can remain part of the wider governance design, but it does not replace a short automatic sunset and urgent independent review. Emergency action should end because the trigger has ended or the clock has run, not because the institution eventually reaches a convenient meeting date.

These safeguards are not procedural decoration. Each corrects a particular incentive created by self-judged urgency. A trigger taxonomy makes it harder to rename ordinary preference as crisis. An evidence burden makes the institution internalise the cost of bypassing discussion. Contemporaneous reasons let outsiders challenge the decision before it hardens. A sunset resists institutional inertia. Independent review separates action from final judgment. Reversibility protects running systems when the initial facts are wrong.

The missed narrowing opportunity

The August proposal should be judged by what it chose to leave intact. The drafters had current Article 11.4 before them. Historical policy materials had already shown that explanation, consultation and limits around Board prerogative were contested. One proposal contemplated explanation and review concepts; AFRINIC staff commentary resisted a two-week consultation constraint as inconsistent with the leeway required for urgent and exceptional circumstances. Those records do not amend the Bylaws, prove an invocation or confer greater authority. They do show that the tension between speed and process was visible years before the 2026 review.

The proposed Constitution nevertheless states that Article 11 has no substantive change. It preserves a clause in which the Board determines necessity and urgency, while the emergency itself remains undefined. It preserves general corporate controls without converting them into a contemporaneous Article 11.4 evidence and disclosure duty. It preserves a schedule-linked later checkpoint without adding a fixed sunset. It preserves the possibility of immediate policy action without naming an independent urgent reviewer.

That is not neutral continuity. In governance, retaining broad discretion after its risks have become clear is a choice. It places the cost of ambiguity on operators and members who must respond to the policy, rather than on the institution that invokes the exception. It invites future directors to resolve missing limits in the heat of an incident, when speed, fear and organisational self-protection are strongest.

AFRINIC needs the capacity to protect unique, accurate shared state. It may sometimes need to move before a long deliberative process ends. But the legitimacy of that movement comes from the narrowness of the technical task, the evidence of a real threat, the preservation of running networks, the reversibility of the response and the availability of independent review. It does not come from the Board’s declaration, a corporate vote or the attendance of a community.

The right constitutional settlement is therefore a thin exception, not a blank cheque: measurable triggers, a public evidence record, minimal containment, conflict disclosure, rapid notice, fixed expiry, independent review and rollback. Anything broader risks laundering an internal coordination role into practical government. The registry remains a bookkeeper even when the clock is running.