Regulatory Capture Without a Regulator

Summary

• A self-governing registry can be captured even though there is no state regulator to capture. The relevant outcome is a repeated redirection of institutional choices from the registry's stated purpose toward a concentrated interest with superior access to money, time, evidence, staff, suppliers or remedies.
• Participation, expertise, employment history and commercial relationships are not proof of capture. A valid diagnosis must connect a mechanism to decisions and then to unequal benefits or burdens while testing alternative explanations such as technical merit, security, lawful compliance and ordinary constituency representation.
• The strongest audit uses several linked records: revenue and fee exposure, board and meeting rosters, proposal authorship, speaking concentration, evidence cited, staff interventions, vendor awards, career movement, implementation burdens, exceptions, complaints and review outcomes.
• Capture can come from several directions. Large members may dominate finance or attendance; professional insiders may control scarce time; staff may shape the feasible agenda; vendors may become indispensable; and a defensive coalition may control both the initial decision and the route for challenging it.
• Openness is necessary but limited public evidence. An open room can still be structurally closed when only a small group can fund repeated attendance, interpret specialised material, produce credible data, survive long consultations and monitor implementation after everyone else leaves.
• The appropriate response is measured insulation rather than exclusion: disclose interests, diversify evidence, publish denominators, separate advice from decision authority, audit vendor dependence, fund underrepresented operators, preserve dissent, and place consequential review outside the original management chain.

Capture begins with a baseline, not a villain

The term capture loses value when it means that an institution listened to a member, employed an industry expert or reached a result that benefited a large operator. Number-resource institutions exist to coordinate with network operators. Their staff need technical experience. Members legitimately finance much of the work. Vendors can supply expertise that would be wasteful to duplicate. A decision that helps an influential constituency may also be the best decision for accuracy, security or continuity.

A serious diagnosis starts with a baseline. What is the institution supposed to do? The answer should come from its charter, articles, service agreements, adopted policy principles and public commitments. Typical purposes include accurate registration, fair access, globally unique delegation, reliable transfer records, technical coordination, member service and continuity. The audit then asks whether decisions are consistently or repeatedly redirected from those purposes toward a narrower interest.

That formulation adapts the OECD's account of policy capture without pretending that a private registry is a government agency. The OECD's 2017 review defines capture by the repeated direction of public decisions away from the public interest toward a specific interest and recommends plural participation, transparency, accountability and organisational integrity. In a registry, the benchmark is not an abstract national public interest. It is the institution's stated number-resource purpose, the legitimate interests it has undertaken to serve and the constraints imposed by applicable law.

Three elements follow. First, there must be directionality: a decision or sequence of decisions moves away from the baseline. Second, there must be concentration: an identifiable interest receives a persistent advantage not adequately explained by the merits. Third, there must be a mechanism: money, access, information, personnel, procurement, agenda control or review explains how influence becomes outcome.

This definition is deliberately demanding. It protects institutions from casual allegations and critics from cosmetic transparency. A roster showing several employees of large networks is not enough. Neither is a board vote, a former staff member changing jobs or a contract with a familiar supplier. The evidence must show a linked pattern across influence, decision and effect.

Self-regulation changes the route, not the possibility

Classic capture theory assumes a public authority whose coercive powers are valuable to regulated firms. A number registry has a different legal form, but its decisions can still be valuable. It maintains recognised records, operates account and transfer services, applies membership conditions, supports reverse DNS and route-origin authorisation, sets charges, and interprets rules that affect how an operator can use or move scarce resources. Dependence on those functions gives interested parties reasons to shape them.

Self-regulation can reduce some capture risks. Entities often possess the knowledge needed to make technically sound rules. Open mailing lists and meetings can expose proposals to direct operational criticism. A membership electorate can remove directors. Distributed communities may resist a single government's political pressure. Public archives can preserve arguments more fully than a closed administrative file.

The same features can create other risks. Expertise is unevenly distributed. Attendance is costly. A policy discussion may be open to everyone while being continuously monitored by only a few paid specialists. The institution that supplies neutral facts may also employ the staff who must implement the result and defend the budget. Members may be both customers and governors. Vendors may advise on the need for services they are later paid to provide. Review may return to people whose prior judgment is under challenge.

The RIPE Accountability Task Force's final report illustrates why formal openness should not end the inquiry. It describes an open community without formal membership, a staffed association that supports the community, archived discussions and a substantial body of written documents. It also notes reliance on norms and identifies areas in the relationship between community and association that were not fully formalised. None of that establishes capture. It shows the institutional terrain in which authority can be distributed through law, convention, expertise and repeated participation rather than one statutory chain.

The central point is functional. Where an institution can confer, withhold, price, delay, certify or review an essential service, an interest may seek durable advantage from that authority. Capture remains possible even when the authority arose by contract, recognition and coordination rather than legislation.

Six candidates can exert concentrated influence

The first candidate is a coalition of large members. Large operators may contribute material revenue, employ dedicated policy staff, produce extensive operational data and attend every meeting. Their scale can improve decisions. It can also allow their cost structure, risk tolerance and commercial strategy to become the unspoken definition of feasibility. Small access providers, new entrants, public networks and operators in high-cost locations may be formally equal yet unable to sustain the same presence.

The second candidate is the professional participation class. This group is not defined only by company size. It includes consultants, association representatives, lawyers, long-serving chairs and people whose employment permits continuous attention. They accumulate procedural memory and social trust. That knowledge is valuable, but repeated control of drafting, agenda setting and consensus interpretation can turn an open forum into a system governed by those who can remain present.

The third candidate is staff. Staff possess implementation knowledge, internal cost estimates, legal advice, incident history and the ability to convert broad decisions into operational practice. A responsible institution needs staff to explain consequences. The risk appears when facts, recommendations and institutional preferences are not distinguishable, or when a community can choose only among options already narrowed by the organisation that will execute them.

The fourth candidate is a vendor or supplier network. A cloud provider, security consultant, election contractor, auditor, software supplier or legal adviser can become difficult to replace. A supplier may influence requirements, frame risk, hold essential knowledge and later win extensions because switching appears dangerous. Capture here does not require bribery. Dependency, information asymmetry and repeated sole-source logic can be enough.

The fifth candidate is a board-management alliance. Directors may rely heavily on management for information, meeting preparation, legal framing and performance measures. Management may rely on the board for strategy, remuneration and institutional protection. A close relationship is not improper, but the combination can reduce independent challenge when the board receives one account of operational necessity.

The sixth candidate is a defensive litigant or pressure coalition. A party with resources to sustain legal threats, public campaigns or repeated complaints can alter institutional risk appetite. The result may favour that party directly, or it may lead the registry to design rules around avoiding a particular kind of challenge. Evidence of pressure alone is not proof; the audit must show that decisions shifted without a reason proportionate to the underlying legal or operational risk.

These candidates can compete. Large members may oppose staff expansion. Staff may resist a dominant vendor. Small operators may align with incumbents on security but not fees. Capture is therefore not a single industry-versus-registry story. It is a changing network of concentrated capacities.

Financial capture is about dependency, not just donations

Membership registries usually do not resemble election campaigns funded by private donors. Their exposure lies in revenue structure, fee sensitivity and expenditure commitments. If a small set of account holders supplies a large share of revenue, the institution may hesitate to impose costs on them. If all members pay a flat fee, a numerically large class may have voting strength even when another class bears most operational impact. If new services create permanent staffing and supplier commitments, the budget itself can make continuation appear necessary.

A financial audit should begin with concentration. The useful measures include the share of revenue attributable to the largest beneficially connected member groups, the proportion of income exposed to a change in one policy, the ratio of restricted to discretionary reserves, and the percentage of expenditure locked into multi-year contracts. Legal entities should be grouped by disclosed control where lawful and proportionate; counting subsidiaries as independent interests can understate concentration.

Budget visibility also matters. The RIPE NCC's 2026 draft activity plan and budget publicly identified a planned EUR 41.1 million income and expense budget, an expected 20,000 LIR accounts, a EUR 1,800 membership fee and divisional changes. Publication allows members to challenge assumptions. A capture audit would go further by tracing which constituencies requested each material activity, who bears its cost, which suppliers benefit, what exit cost has accumulated and whether reported outcomes match the original justification.

The direction of influence cannot be inferred from payment alone. A high-paying operator may receive no special treatment. A low-paying but numerous group may control an election. A service line may benefit the whole routing system even if its direct users are few. The relevant question is whether financial dependence predicts decisions after technical need, legal duty, service demand and stated policy goals are considered.

Useful warning signs include repeated exemptions for revenue-critical groups without equivalent published reasoning; budgets that expand around services promoted by the same suppliers who deliver them; consultation choices framed as accepting one expenditure path or risking continuity; and fees whose burdens are assessed by legal account rather than actual controlling group. None is conclusive alone. Together with outcome evidence, they identify where deeper review belongs.

Time is a governance currency

Open participation is often defended by pointing out that anyone may join a list, attend remotely or speak at a microphone. That describes access, not usable influence. Participation requires attention before the meeting, familiarity with prior versions, ability to interpret specialised claims, confidence to intervene, and time after the meeting to monitor drafting and implementation. These resources are unequally distributed.

The audit should therefore measure participation in hours and continuity, not only headcount. For each material decision, record unique entities, repeat entities, first-time entities, speaking time, proposal authors, editors, chairs, review commenters and affiliations. Measure how many organisations supplied people at three or more stages. Compare the attendance denominator with the membership and with the operators materially affected. Count not only supportive interventions but substantive objections and whether they received an answer.

Professional time can create what looks like consensus. If hundreds of eligible members remain silent while twelve specialists discuss a proposal for months, the evidence establishes agreement among the active group, not necessarily the preferences of the full constituency. That active agreement may still be the best available decision rule. The problem begins when silence is narrated as affirmative mandate or when the institution does not disclose the denominator.

Participation grants, remote access and translated material can lower barriers, but they should be evaluated by effect. Did new operators author text, introduce evidence, change a proposal or serve in decision roles? Attendance photographs and registration counts do not answer those questions. Nor does geographical diversity automatically produce economic or operational diversity if entities depend on the same employers, sponsors or professional network.

The capture signal is a stable concentration of decisive labour. If the same affiliated cluster repeatedly defines problems, drafts solutions, interprets objections and confirms completion, while affected outsiders enter only after implementation, procedural openness has not produced plural control. The remedy is not to exclude experienced entities. It is to make their concentration visible and supply countervailing capacity.

Information can capture before a vote occurs

Number-resource decisions depend on specialised evidence: allocation and transfer records, routing measurements, abuse reports, certification incidents, legal constraints, service costs and implementation estimates. The party that defines the evidence can define the range of plausible choices. Information capture occurs when decision makers become dependent on a narrow supplier of facts and lack the capacity to verify the supplier's assumptions.

Staff often hold the richest information because they operate the services. Large networks can provide traffic and deployment experience unavailable to smaller members. Vendors know their products. Lawyers understand litigation exposure. Each source is legitimate. The governance risk lies in unlabelled roles and missing counterevidence. A cost estimate prepared by the unit seeking budget approval is not automatically wrong, but it should not be treated as independent validation.

For every consequential proposal, an evidence register should identify each empirical claim, its provider, relevant affiliation, date range, denominator, method, limitations and whether underlying material can be reviewed. Claims that cannot be published for security, privacy or legal reasons can be examined by an independent reviewer under controlled access. The public result should still explain what category of evidence was checked and what uncertainty remains.

Information capture can be measured. Calculate the share of cited empirical claims supplied by the institution, the largest members, one vendor or one professional network. Track whether contrary datasets were requested. Compare forecasts with realised outcomes. Record how often later implementation reveals costs or affected groups absent from the original analysis. Repeated one-sided forecasting followed by unexamined renewal is a stronger warning than a single imperfect estimate.

The diagnostic must also guard against performative evidence. A long technical paper can dominate a discussion without answering the decision question. A dashboard can display precision while omitting the population most burdened. An incident anecdote can justify a broad restriction without showing frequency or causation. Evidence quality depends on relevance, denominator, reproducibility and explicit uncertainty, not volume.

Staff authority requires role clarity

Self-governing communities need professional staff. Volunteers cannot safely operate continuous registry, security and member services by themselves. Staff preserve institutional memory, warn about implementation hazards and provide facts across changes in community participation. Treating expertise as suspect would weaken the institution and make it more dependent on external suppliers.

Role confusion is the risk. Staff can be fact providers, advisers, drafters, facilitators, implementers and affected employees in the same debate. If a proposal changes staffing, service scope, procurement or management discretion, the organisation has an interest in the outcome. That interest does not disqualify its evidence, but it should be disclosed and separated from the authority to declare consensus or approve the result.

The RIPE community's 2023 statement on staff participation addresses this tension directly. It welcomes the expertise of RIPE NCC staff while stating that, where the community provides guidance to the RIPE NCC, staff should disclose their position and avoid giving direction. The text is not evidence that staff captured any decision. It is evidence that the dual role is real enough to require an explicit boundary.

An audit can test that boundary by coding staff interventions. Was the contribution factual, interpretive, recommendatory or determinative? Did staff provide alternatives, costs and uncertainty, or only the preferred path? Who drafted the final wording? Who decided whether objections were resolved? Did a board or community body receive independent advice where management's own authority or budget was at issue?

The same method applies beyond RIPE. It does not assume all registries share one legal or community structure. It asks a common functional question: when the organisation receiving guidance also supplies expertise, how is self-direction prevented? A strong answer combines role labels, published impact analysis, independent evidence, chair authority outside management and later comparison between forecast and result.

Revolving doors need outcome evidence

Career movement between registries, members, vendors, law firms and peer institutions is predictable in a specialised field. A blanket ban would deprive institutions of scarce competence and punish normal professional development. Employment history should therefore be treated as a risk indicator, not a verdict.

Empirical research shows why both attention and restraint are warranted. A study of United States patent examiners found that examiners granted more patents to firms that later hired them, with effects strongest around active hiring; the authors report the results in the NBER working paper on patent-examiner mobility. That finding demonstrates a measurable route in one agency. It does not prove that similar movement in an Internet institution produces the same effect.

Other research warns against diagnosis by reputation. A detailed study of alleged pharmaceutical capture tested revolving-door, information-overload and shared-culture mechanisms and found limited support in the examined case; its authors emphasise the danger of treating a scandal label as evidence, as described in the published study's repository record. The lesson for registries is to specify the predicted trace before examining names.

The register should cover current employment, recent employment, material clients, beneficial interests, board roles, vendor relationships and litigation positions. Cooling-off rules should be tailored to decisions, not merely elapsed time. A former registry employee advising a member on an unrelated technical issue is different from negotiating a contract they designed. A director employed by a large operator may contribute valuable knowledge but should not decide a matter uniquely benefiting that employer without disclosure and recusal.

The decisive analysis links movement to outcomes: unusually favourable timing, exceptions, specifications tailored to a future employer, procurement access, confidential knowledge or repeated decisions benefiting connected parties. It should compare similarly situated applicants or proposals. Without that comparison, a career map shows proximity, not capture.

Vendor capture grows through switching costs

Suppliers can gain institutional power even when procurement was fair. Custom software, identity systems, security operations, election platforms, external counsel and specialist audits accumulate knowledge. Over time, replacement appears risky, internal staff adapt to the supplier, and future requirements are written around the installed service. The vendor becomes part of the institution's definition of continuity.

The audit should map the full dependency. What operational knowledge resides only with the supplier? Which interfaces are portable? Can records and configurations be exported in usable form? How many contract extensions occurred without competition? Did the supplier advise on the need, write technical requirements, assess alternatives or evaluate its own performance? What percentage of the institution's critical functions depends on one corporate group?

Price is an incomplete measure. A low-cost vendor may hold severe exit leverage. A high-cost adviser may be readily replaceable. The strongest indicators are untested recovery, proprietary integration, concentration of administrator access, weak documentation, absent successor rights and repeated claims that competition would endanger service. An independent continuity exercise can reveal more than a procurement compliance checklist.

Vendor influence can also travel through language. A supplier's risk taxonomy may become the board's taxonomy. Product capabilities may define strategic priorities. Metrics may report inputs the vendor can easily count rather than outcomes members need. Reviewing the origin of requirements and performance indicators is therefore part of the capture analysis.

The remedy is not automatic insourcing. Internal monopolies can be less transparent than external contracts. The answer is contestability: modular interfaces, documented configurations, audit access, assignment rights, data portability, key-person controls, competitive benchmarks and rehearsed transition. A supplier that performs well should be capable of winning renewal without institutional dependence doing the deciding.

Governance documents show safeguards, not immunity

Conflict rules, elections, annual reports and open meetings are material protections. They should be credited. They do not prove that capture is absent, because formal compliance can coexist with concentrated agenda power, information dependence or unequal participation.

The RIPE NCC Articles of Association provide that an Executive Board member with a direct or indirect personal conflict should not participate in the relevant deliberation and decision. They also give the General Meeting roles in reporting and corporate decisions. Those rules create observable events: disclosure, recusal, alternative representation and member action. An audit can test whether the rules are triggered consistently and whether the published record is sufficient to understand treatment.

ICANN offers another relevant comparator. Its Board Conflicts of Interest Policy calls for annual disclosures, review of potential conflicts, records of affected persons, the nature of a conflict, deliberation and votes. Its Board Code of Conduct links ethical risk, accountability and fair disclosure. ICANN governs domain-name coordination rather than regional number registration, so its instruments cannot simply be copied. They demonstrate that technically specialised private governance can document conflict handling in detail.

The test is effectiveness. Did conflicted persons leave the relevant discussion? Were indirect interests captured? Could the remaining decision makers obtain independent information? Was the reason for the decision preserved? Were supplier and employment conflicts included, or only direct financial holdings? Did the safeguard cover staff, chairs and reviewers as well as directors?

A policy can be exemplary on paper and weak in operation. Conversely, a community with modest formal language may enforce strong norms. The diagnosis should compare written safeguards with observed decisions and avoid rating institutions by document length.

A capture audit needs a linked evidence model

The minimum unit of analysis is a decision episode. It begins when an issue enters the agenda and ends after implementation, challenge and review. For each episode, the audit should preserve the baseline purpose, decision makers, active entities, affiliations, evidence providers, financial exposures, suppliers, options considered, reasons, votes or consensus finding, implementation effects, exceptions and remedies.

Network analysis can then reveal relationships invisible in isolated disclosures. Nodes include people, employers, controlled corporate groups, vendors, committees and decisions. Edges include employment, payment, board service, proposal co-authorship, sponsorship, recurring meeting participation, procurement, advisory work and review roles. Every edge needs a date range and source. Historical ties should decay in weight unless continuing relationships justify otherwise.

Several measures are useful. Degree concentration shows actors with many direct connections. Betweenness identifies people who bridge otherwise separate groups and may control information flow. Affiliation assortativity shows whether debate clusters by employer or commercial interest. Repeated co-authorship can reveal a stable drafting coalition. Decision-to-benefit edges show which groups repeatedly gain exemptions, lower costs, faster service or greater authority.

Numbers require interpretation. A staff secretariat will naturally appear central because it publishes material and attends meetings. A chair will connect many entities by role. A dominant transit provider may be central because it has genuine operational reach. The audit should compare observed centrality with role-expected centrality and focus on unexplained influence over contested outcomes.

Time order is essential. A relationship formed after a decision may be relevant to revolving-door analysis but cannot be treated as prior lobbying without evidence. A vendor contract preceding a requirement change raises different questions from one awarded after open competition. A entity's affiliation at the time of each intervention matters more than their present biography.

The model should support challenge. People named in a risk review need an opportunity to correct affiliations and provide alternative explanations. The objective is institutional learning, not a social graph of suspicion.

Nine indicators turn concern into a reproducible test

The first indicator is financial concentration: the share of revenue, threatened withdrawal, fee benefit and contractual expenditure connected to the interest. The second is participation concentration: speaking, drafting, chairing and review work performed by that interest across the full episode. The third is information concentration: the portion of decisive claims and forecasts supplied or controlled by connected actors.

The fourth is agenda control: who can place, defer, narrow or remove an issue. The fifth is personnel circulation: whether relevant decision makers, advisers and beneficiaries moved between connected roles before or after the decision. The sixth is supplier dependence: whether the institution can realistically reject or replace a vendor without unacceptable continuity loss.

The seventh is outcome alignment: whether the decision repeatedly benefits the interest compared with similarly situated groups. The eighth is burden displacement: whether costs, delays, security exposure or loss of portability fall on less represented operators or the wider network. The ninth is remedy control: whether the same connected group influences the initial decision, internal reconsideration and external response.

Each indicator should be rated for evidence quality as well as risk. A high-risk claim based on an anonymous anecdote is not equivalent to one supported by contracts, timed affiliations and comparative decisions. One practical scale is zero for no observed signal, one for a plausible but weakly evidenced signal, two for a documented concentration without demonstrated outcome effect, and three for a documented mechanism connected to repeated outcome and burden. Evidence quality can be separately rated from zero to three.

No total should automatically declare capture. Aggregation can hide the causal chain. Instead, require at least one strong influence mechanism, one strong outcome indicator and evidence that competing explanations were tested. Financial concentration plus vendor dependency, for example, is not capture unless it helps explain institutional choices. Outcome alignment alone is also limited public evidence if the favoured result follows technical necessity.

The framework should publish both positive and negative findings. If a well-connected member lost a decision on the merits, that is counterevidence. If staff proposed several options and the community selected one staff did not prefer, that weakens a staff-capture hypothesis. A diagnostic that records only suspicious facts will manufacture its conclusion.

Counterfactuals separate influence from capture

Every capture claim needs a counterfactual: what would a decision more faithful to the institution's stated purpose probably have looked like? The answer cannot be whatever the critic preferred. It should be derived from comparable cases, published criteria, prior practice, independent evidence and the treatment of similarly situated parties.

Suppose a policy increases transfer verification costs and large members support it. The capture hypothesis predicts that incumbents use compliance costs to disadvantage smaller entrants. The security hypothesis predicts that fraud evidence justifies stronger verification and that burdens track risk. The audit would compare fraud rates, cost distribution, available lower-burden controls, exception design, implementation timing and outcomes after adoption. If costs fell mainly on small operators without measured security benefit and alternatives were dismissed without analysis, capture becomes more plausible.

If fraud declined materially and assistance reduced entry burdens, the security explanation strengthens.

Suppose staff recommend expanding a registry-operated service. A bureaucratic-expansion hypothesis predicts selective cost estimates, dismissal of external alternatives and success measures based on staffing or activity. A public-goods hypothesis predicts unmet member demand, benefits that private suppliers would underprovide, transparent alternatives and outcome measures tied to network value. Later performance can distinguish them.

Suppose a long-serving vendor wins another contract. A vendor-capture hypothesis predicts requirements tailored to the incumbent, absent portability tests and exaggerated transition risk. A competence hypothesis predicts open comparison, documented switching costs, strong performance and credible exit rights despite renewal. The same outcome can arise from different mechanisms.

Counterfactual discipline also protects minorities. A proposal should not be condemned because a majority supported it, nor validated because an underrepresented group opposed it. Capture concerns the relationship between concentrated influence and deviation from purpose, not the identity of the winner.

Remedies must preserve expertise while reducing dependence

The first safeguard is a structured interest register covering employment, controlled companies, material clients, board roles, vendors, litigation and recent institutional positions. Disclosure should be event-sensitive: a relationship immaterial to one decision may be decisive in another. Recusal records should identify the stage from which a person withdrew and how independent expertise was replaced.

The second is a participation denominator. Every consequential decision should report eligible members or affected operators, unique entities, organisational affiliations, first-time contributors, substantive objectors, proposal authors and active reviewers. Consensus language should describe the observed group accurately rather than converting absence into endorsement.

The third is independent evidence capacity. A modest standing budget can commission validation of cost, security and market claims when the institution, a large member or a vendor supplies the decisive analysis. Independence requires control over questions, access to necessary material, disclosure of assumptions and publication of limitations, not merely a different letterhead.

The fourth is role separation. Staff may explain facts and implementation, but a chair outside management should determine community agreement. A board committee should obtain independent advice when management authority, remuneration or procurement is at issue. Reviewers should not report through the chain responsible for the original act.

The fifth is contestable procurement and operational portability. Contracts should include documentation, export, audit, transition assistance, credential transfer and successor rights. Periodic recovery exercises should test whether these rights work. Vendor performance and dependence should be reported together.

The sixth is countervailing participation. Travel support, remote facilities, translation, research grants and paid release time can help smaller operators contribute. Selection should be transparent and should not make recipients dependent on supporting the institution. Success is measured by substantive contribution and influence, not attendance.

The seventh is a protected dissent record. Final decisions should state material objections, evidence considered, uncertainty and reasons for rejecting alternatives. Later reviewers can then see whether an ignored warning became real. Institutional memory is an anti-capture control because it prevents the prevailing coalition from rewriting the history of choice.

Review is the decisive test of institutional independence

An institution may tolerate criticism at the consultation stage yet remain captured if challenge after the decision is controlled by the same network. Remedy control deserves special weight because it determines whether error can be corrected when influence becomes visible.

The audit should map who receives a complaint, who selects reviewers, who pays them, who supplies the record, who can pause an irreversible act, who writes the response and who implements correction. Formal separation is weak if reviewers depend on management for reappointment or see only a summary prepared by the original unit. Independence must cover information, budget, tenure and remedial authority.

Outcome statistics are necessary but easily misread. A low success rate may show sound initial decisions or ineffective review. A high success rate may show accessible correction or poor first-instance quality. Publish time to resolution, interim protection, grounds raised, correction type, repeat issues and implementation. Where confidentiality applies, aggregate results can still reveal patterns.

Capture risk rises when connected parties receive informal resolution while outsiders face formal delay; when exceptions are granted without published criteria; when the institution changes reasons during review; or when the reviewer can recommend but not secure correction. Comparative treatment matters more than raw win rates.

External courts remain important where legal rights are engaged, but litigation cost and delay make them an incomplete governance safeguard. Internal reconsideration can correct routine errors. Independent panels can address specialist disputes. Courts can determine legal authority. A credible system distinguishes these functions instead of presenting any one as complete immunity from capture.

A hypothetical audit shows what proof looks like

Consider a fictional registry proposal to require a new automated verification service for every transfer. The proposal is authored by two long-serving entities employed by large brokerage-adjacent firms. Registry staff provide fraud figures, and the only cost estimate comes from the existing identity vendor. Smaller operators entity that the service adds fixed costs and excludes documents common in several jurisdictions. The board approves a multi-year vendor extension after the policy is accepted.

This sequence raises several indicators: concentrated authorship, one-sided evidence, vendor dependence, burden on smaller operators and linked procurement. It still does not establish capture. The audit must test whether transfer fraud was material, whether the authors' employers benefit, whether alternative controls were considered, whether the vendor's estimate was independently checked, whether the board procurement followed fair requirements and whether the system improved outcomes.

Imagine the deeper record shows that fraud had risen sharply across operator sizes; the policy added a low-cost manual route for difficult jurisdictions; an independent security team validated the control; the vendor won against credible alternatives; and performance review after one year showed fewer fraudulent transfers without disproportionate delay. The initial appearance of capture weakens substantially.

Now change the facts. Fraud data combine unrelated account incidents, the vendor drafted the risk definition, no independent estimate exists, manual review is available only to the largest members, the connected firms receive volume discounts, and complaints return to the unit that designed the service. After two years, delays rise while fraud outcomes are not published. The linked evidence now supports a much stronger diagnosis.

The example demonstrates why the unit of proof is not a relationship. It is a causal chain: concentrated capacity shaped the evidence and options; the decision advantaged connected interests; burdens moved to weakly represented parties; predicted public benefits were unverified; and review could not correct the design.

Publication should expose patterns without exposing people unfairly

A capture review must be transparent enough to support trust and careful enough to avoid reputational punishment by association. Public reports should distinguish verified fact, analytical inference and unresolved allegation. Affiliations should be dated. Financial estimates should use ranges where exact disclosure would expose confidential member information. Private identities unrelated to decision authority should not be published merely because they appear in correspondence.

Institutions should offer correction before publication. Affected people can identify outdated employment, misunderstood control relationships or missing recusals. Their response should not give them a veto over findings. Material disagreements can be presented with the basis for the final assessment.

Risk ratings should apply primarily to decisions and institutional arrangements, not character. Saying that a procurement episode had high vendor-dependence risk is more accurate than declaring a supplier corrupt. Saying that a policy episode relied on a concentrated entity network is more useful than suggesting every entity acted in bad faith.

The report should also identify strengths: diverse evidence, effective recusal, a decision contrary to a dominant funder's preference, successful appeal or tested vendor transition. Capture prevention depends on learning which controls work, not only naming vulnerabilities.

Regular publication creates a time series. One year may show concentration because an emergency required specialists. Five years may reveal whether the same cluster remains decisive across unrelated issues. Repeated measures make institutional drift visible without presuming that continuity is itself improper.

The evidence standard should be high because the stakes are high

Capture allegations can delegitimise necessary coordination. They can be used by disappointed parties to avoid an adverse rule, by incumbents to attack reform, or by political actors seeking control of a private institution. The standard of proof must therefore reject guilt by employment, expertise, nationality, attendance or commercial success.

At the same time, demanding direct evidence of a secret bargain would make capture nearly impossible to diagnose. Modern influence often operates through legal participation, shared assumptions, dependency and unequal capacity. The appropriate standard is convergent evidence: several independent records support a specified mechanism, observed outcomes align with its predictions, and credible alternatives fit less well.

The research on alleged capture supports this caution. Career links may predict favour in one setting and fail to explain outcomes in another. Open participation may broaden decisions in one institution and merely legitimise insider control in another. Conflict policies may prevent direct self-dealing while leaving information or agenda concentration untouched. Context determines the trace.

A high-confidence finding should therefore answer six questions. What baseline purpose was displaced? Which concentrated interest benefited? What mechanism converted capacity into influence? Which decisions and effects demonstrate repetition? What competing explanations were tested? Which remedy or institutional change would reduce the mechanism without destroying useful expertise?

If any answer is missing, the result should be a risk finding or an unresolved hypothesis, not a declaration of capture. Precision is part of accountability.

A practical annual capture statement

Each registry could publish an annual statement organised around decisions rather than public-relations themes. It would list consequential policy, fee, service, procurement and review episodes; participation denominators; material interest disclosures and recusals; sources of decisive evidence; forecast and realised effects; supplier concentration; exceptions; complaints; and corrective action.

The statement should include a concentration table without exposing confidential customer details. Revenue share can be grouped by beneficially connected cohort. Participation can be grouped by employer type and geography. Evidence provenance can distinguish staff, member, vendor, independent and public sources. Procurement can report competition, extensions, exit testing and critical dependencies.

It should also report questions that could not be answered. Missing beneficial-control information, absent meeting affiliation data, confidential settlements or unmeasured implementation burdens are findings because they limit assurance. The institution can then improve the next year's record.

Independent reviewers should sample episodes rather than certify the whole institution with a generic clean opinion. Selection can combine high-impact decisions, random decisions and episodes flagged by indicators. Review terms and funding should be fixed before cases are chosen. The reviewer should be able to inspect protected material and publish a reasoned conclusion.

Members and the wider operator community should discuss the statement, but acceptance should not depend only on those whose influence it evaluates. An external governance committee or rotating cross-regional panel can challenge methods while leaving substantive number-resource authority where it belongs.

The annual statement is not a score of virtue. It is a maintenance record for institutional independence.

Capture prevention is a design duty, not an accusation

Self-governing number registries have real advantages. Technical knowledge sits close to implementation. Operators can challenge impractical ideas. Regional communities can adapt general coordination to different conditions. Membership funding can sustain public goods. These strengths should not be sacrificed to a fantasy of governance without interested entities.

But self-government does not dissolve power. It redistributes power among members, staff, boards, professional entities, suppliers and reviewers. Where participation costs, information and switching barriers are unequal, an apparently open institution can become dependent on a narrow group without any illegal bargain and without one dramatic takeover.

The right response is to observe the routes by which influence becomes durable. Follow revenue, but also time. Follow declarations, but also evidence. Follow employment, but also decisions before and after movement. Follow vendor price, but also exit capacity. Follow votes and consensus, but also denominators, burdens and remedies.

No single measure establishes capture. A linked and repeated pattern can. When the institution publishes enough information to test that pattern, it protects legitimate expertise from suspicion and exposes dependency before it becomes constitutional. When it refuses measurement and relies only on the words open, bottom-up or member-led, it asks the public to treat institutional form as proof of institutional independence.

Capture without a regulator is therefore neither paradox nor slogan. It is the possibility that a private coordination body repeatedly serves a concentrated interest because that interest controls the resources of governance. The safeguard is not hostility to members or staff. It is a measurable distribution of money, knowledge, voice, decision authority and correction strong enough that no one constituency can quietly become the institution itself.