ROL daily
Measuring Rule of Law through documentary evidence — now available open access. → Read the paper· Bengio names the risk. The SROL names the standard. Read →·Fable 5 ban: no criteria, no standard, pure discretion. Read →·SCOTUS immunity doctrine: lines drawn in the wrong dimension entirely.·AI in governance is an exercise of state power. The ROL standard applies now.·Moratorium or standard? The ROL has always had the answer. Read →·The next big thing in governance is the oldest framework, finally made precise. Read →·

Category: The Missing Foundation: Why Current Proposals for Governing AI Deployment Ignore Coordination Architecture

Overview and critique of policy and other documents issued by major AI providers on bridging AI and governance

  • There’s a Hole in the Bucket

    Anthropic’s Framework governs the most dangerous AI. It is not governing the most common AI use case.

    Rita Felgate | August 2026
    ruleoflaw.science

    An em-dash publication: developed through human-AI cognitive synthesis with Claude (Anthropic). Learn about the em-dash methodology


    Something is missing from every serious AI governance framework currently in circulation. Not from malice. Not from oversight. The frameworks are addressing the risks they were designed to address. But the most common AI use case — AI drafting, processing, and operationalising the instruments of state power — is not in any of them.

    That is the hole in the bucket.


    The Framework and the Use Case It Does Not Address

    Anthropic’s Advanced AI Framework, published June 2026, is serious work. It addresses catastrophic risks from frontier AI — biological weapons, offensive cyber operations, loss of control of AI systems, and automated research and development. It proposes mandatory independent evaluation, transparency obligations, and enforcement authority. It is the kind of framework that governance of dangerous AI requires.

    It is entirely silent on the most common AI use case.

    The AI that will affect most people most often is not the frontier model designing bioweapons. It is the AI system drafting the regulation that sets the fine for your expired licence, processing the ministerial notice that determines your building permit, generating the commencement notice that brings an Act into force — or fails to. Blog 2 in this series showed what the Rule of law requires of AI when it enters that chain. Every one of those instruments must satisfy the Seven Rule of law (ROL) Compliance Categories — Authority, Jurisdiction, Clarity, Public Participation, Publication, Referent, Commencement. These conditions are necessary to the coordination capacity of the state and hence its stability. Anthropic’s Framework addresses none of this.

    This is not a criticism of Anthropic. It is an identification of a different problem — one the Framework was not designed to solve and cannot solve without a measurement standard it does not currently have.


    Independent Evaluation Needs a Standard

    The Framework proposes mandatory independent evaluation of AI systems. That is the right instinct. But the Framework scopes its evaluation to four Enumerated Risks — biological weapons, offensive cyber operations, loss of control of AI systems, and automated research and development. None of those is the documentary compliance of governance instruments. The Framework itself says it wants evaluator standards published. For the use case this blog addresses, that standard does not yet exist.

    Consider the structure of any governance preservation infrastructure. It records which standard was applied and what result followed. What it cannot supply is the standard itself — the observable conditions for ROL-compliant coordination that any governance instrument must satisfy. Preservation without an upstream standard produces a record of whatever was submitted, compliant or not. The standard must come from upstream.

    Without an upstream standard, third-party evaluation of AI governance instruments produces accountability infrastructure without accountability content. The evaluation is preserved. What it evaluates against remains undefined.

    The Seven ROL Compliance Categories are that upstream standard — published open access, empirically grounded against 28,337 gazette instruments (Felgate, R. (2026). DOI: 10.5281/zenodo.21134975), available today. A third-party evaluator applying the Seven Categories to AI-generated governance instruments produces a verifiable finding: pass or fail on each observable condition, overall verdict, identified coordination defects. That is not another opinion. That is a measurement.


    Government Contract Termination Without Documentary Specification Is Not Governance

    Several European governments have restricted or ended contracts with AI suppliers in governance functions — France, Germany, Spain among them. The stated reasons are supplier-relationship risk: dependence on partners capable of cutting access to critical technology at will. Those states can exit a supplier on those grounds. What they cannot do is state what documentary compliance standard the next supplier must meet. That is the governance gap. A state can name the relationship risk. It cannot yet name the Category failure.

    Terminating a contract without specifying which documentary conditions the supplier failed to satisfy is reaction, not governance. ROL-compliant governance depends on prospective specification of the conditions that make AI’s exercise of state power legitimate — not retrospective termination when something feels wrong. A state that cannot name the Category failure cannot govern the supplier.

    The ROL due diligence clause — introduced in Blog 2 in this series — gives governments the prospective specification they need. It requires AI suppliers to verify and document Seven Category compliance at the point of production. It makes the documentary basis of AI’s exercise of state power visible before the contract ends, not after.

    Governments are currently at the stage of recognising the problem. The measurement standard that lets them name it precisely already exists.


    The First Step Anthropic Can Take Today

    In June 2026, California made Claude the first AI tool available to all its state agencies and local governments at half price. The Governor’s office stated explicitly that Claude will help state workers draft documents and analyse information. That is the SROL’s exact use case, named by the customer, in a live Anthropic contract, at a scale where a pilot clause is actually negotiable.

    The ROL due diligence clause exists the moment two parties agree to it. It does not require new legislation. It does not require a new regulatory body. It requires an AI supplier willing to warrant the production-side Categories — Clarity, Publication, Referent — and a government willing to warrant the process-side Categories — Authority, Jurisdiction, Commencement — with both parties verifying against the documentary record that Public Participation occurred. The allocation of Categories between supplier and state is developed in Blog 2 in this series.

    California is drafting documents with Claude today. The clause that makes that activity verifiably ROL-compliant could be in the next contract.


    Why This Matters for the Framework Itself

    Anthropic’s Framework is designed to prevent catastrophic risks from AI. But those catastrophic risks do not arise in a governance vacuum. They arise in governance systems whose information architecture is already compromised — where the instruments of state power systematically fail the Seven ROL Compliance Categories, where coordination capacity is degraded, where the state’s ability to detect and respond to emerging risks is already weakened by years of accumulated Category failures.

    A state whose governance instruments fail the Seven Categories is a state whose incident reporting mandates, independent evaluation requirements, and enforcement mechanisms are themselves built on a defective documentary foundation. The Framework’s catastrophic risk governance depends on governance instruments that must themselves be ROL-compliant to function. Securing the ordinary end of AI governance is the foundation that makes the catastrophic end governable.

    The Framework governs the most dangerous AI. One clause governs the most common AI. Both are necessary. One already exists.


    The Seven ROL Compliance Categories are published open access and available for any AI supplier or government to apply today. If the standard your framework is missing is the one described here — get in touch.


    The Science-based Rule of law framework has been in development for twenty years. Its systematic documentary measurement covers sixteen years of South African government gazette data, applying formal scientific methodology to governance. The Seven ROL Compliance Category methodology is published open access: Felgate, R. (2026). DOI: 10.5281/zenodo.21134975.

    Rita V. Felgate is an independent legal practitioner and governance researcher. She is the founder of ruleoflaw.science and the developer of the Science-based Rule of law framework.

  • The Public as Participant, not Audience

    Why the vote needs the same standard as AI — and what that standard reveals

    Rita Felgate | August 2026
    ruleoflaw.science

    An em-dash publication: developed through human-AI cognitive synthesis with Claude (Anthropic). Learn about the em-dash methodology


    When the vote doesn’t work, it doesn’t work for anyone.

    Not for the side that won — because a win that cannot be verified is a win that can always be questioned. Not for the side that lost — because a loss produced by instruments nobody checked is a loss that will always be contested. The vote is supposed to be the moment when the Public’s participation in governance is most visible, most direct, most consequential. Instead it has become the moment when the absence of an observable standard is most exposed.

    That standard exists. It is the same standard that the previous post showed must be satisfied when AI enters the governance chain — the Seven Rule of law (ROL) Compliance Categories, observable, verifiable, already published. It applies to every instrument of state. Including the instruments governing the vote.

    And applying it reveals something the model of governance in general use has never been able to see.


    Three Branches — and the Loose Ends They Leave

    The current governance model — conventionally associated with the Rule of law and with democratic governance — has three branches: Legislature, Executive, Judiciary. This is supplemented, in thicker definitions of the ROL, with matters that evoke the Public, such as rights, freedoms, and participation. Constitutional theory, rule of law scholarship, the separation of powers — all of it built on three branches. The three-branch model is not merely a description. It is the framework that has shaped how governance is understood, how courts review it, how those subjected to governance challenge it. Its authority runs deep.

    And yet it has been missing something.

    The effect? The three-branch model leaves the Public’s entire portfolio of functions without a framework. Each instrument produced in the exercise of the Public’s coordination function — whether produced by the Public, depended upon by the Public, or required of the Public in its relationship to the state’s exercise of power — must satisfy all Seven ROL Compliance Categories — Authority, Jurisdiction, Clarity, Public Participation, Publication, Referent, Commencement. Under the three-branch model, all seven of these are loose ends for the Public’s portfolio. There is no framework to tuck them in.

    Instead, they are tucked into the language of constitutionalism and democracy — normative frames that name the problem without measuring it. The thick rule of law — which fills the gap with democratic values rather than observable coordination conditions — is a product of exactly this exclusion. The Public’s Category failures become “democratic deficits.” The Public’s Jurisdiction failures become “constitutional crises.” The Public’s Commencement failures — the instruments governing its participation that were never properly brought into force — become invisible entirely. The gaps are not resolved. They are renamed.

    The SROL’s measurement instrument does not rename them. It measures them. And what the measurement reveals is that the Public’s coordination function has been systematically unmeasured — not because the function does not exist, but because the model that should account for it has no place for it.


    What the Three-Branch Model Cannot See

    Under the three-branch model, the Public’s functions and operations in governance are definitionally invisible. A model that does not formally recognise the Public as a participant in governance cannot define what the Public does there, cannot identify its outputs, and cannot test whether those outputs satisfy the observable conditions for valid coordination. The Public’s access to and exercise of state powers is simply not in the frame.

    This is not incidental. It is structural. A model that counts three arms of government and stops there has no mechanism for recognising a fourth — and no mechanism for measuring what it cannot recognise. The consequence is that the myriad ways to remove the Public’s coordination function from the chain pass undetected. An uncommenced electoral regulation. A campaign finance rule that was never consulted on. A participation requirement published in a form the Public cannot access. Each one is a Category failure. Under the three-branch model, none of them register as governance failures at all — because the model has no account of what the Public’s coordination function is, and therefore no standard against which its failure can be measured.

    Recognising the Public makes the structural pattern visible. The Gauteng Public Library and Information Services Bill — the statute meant to establish the Public’s own access to information infrastructure — was passed by the provincial legislature in November 2014, and the Speaker’s certified copy was gazetted that December (Gauteng Provincial Gazette No. 377, 25 December 2014), under a heading calling it an Act. The Premier never assented to it. Under sections 121 and 123 of the South African Constitution, without assent there is no Act. In 2018 the MEC published draft regulations under sections of an instrument that had never become law. Eleven years on, assent has still not been given. Asked for the commencement notice, a law library produced the gazette carrying the Speaker’s certified copy: the institutional understanding of what commencement requires is itself defective. That is what the measurement instrument returns. The gazette record shows no assent and no commencement — and the record of the assent that never happened sits where no member of the Public would think to look. The invisibility is not a gap in the model. It is the model operating within its own definitions.

    The SROL’s four-component model changes this. The Public is not the audience for governance. It is a fourth component of state — separate in function, equal in the coordination chain. Separate because the Public’s function is distinct from the Legislature’s production of instruments, the Executive’s exercise of powers under them, or the Judiciary’s adjudication of disputes arising from them. Equal because without the Public’s participation condition satisfied, governance instruments cannot coordinate a ROL-compliant effect with the population they bind. Each component’s coordination function includes a check on the others — the Judiciary reviews the Executive, the Legislature holds the Executive to account, the Public checks each of the three through the documentary record of their instruments. The Public’s check is the instance the three-branch model has no account of, and no mechanism for protecting. That is Category 4 of the Seven ROL Compliance Categories — Public Participation — an observable coordination condition, one of seven, and all seven apply to the Public’s portfolio of functions.

    When the Public is recognised as a fourth component with a defined coordination function, what was invisible becomes measurable. The uncommenced electoral regulation fails Category 7. The unconsulted campaign finance rule fails Category 4. The inaccessible publication fails Category 5. Category 6 asks whether the instrument correctly identified the enabling instrument that confers the power it exercises. An electoral regulation that cites a section repealed before the regulation was made fails Category 6 — the power may exist, but the instrument did not correctly name its source. Each failure is identifiable in the documentary record. Each one is a measurable coordination defect — not a political grievance, not a democratic complaint, but an observable failure of the instrument to satisfy the conditions that make valid coordination possible.


    The Participation Condition Under Attack

    In states where the Public has been largely removed from the coordination chain — by force or by systematic information failure — the Public is silenced. The instruments that carry the conditions for the Public’s participation — election laws, campaign regulations, access to information instruments — fail Categories at the point of production. The Public cannot participate adequately because the information architecture that makes participation possible has been destroyed or captured.

    The same structural problem is now visible where it was least expected: the participation condition being degraded from within, through the information architecture the vote depends on. Not by smashing clay tablets as in antiquity. Not by declaring martial law. By the systematic production of governance instruments that fail against the Seven ROL Compliance Categories.


    Neutered by Information Failure

    The mechanism through which the Public is excluded from the coordination chain is rarely force. It is information failure — the systematic failure of governance instruments to satisfy the conditions that make the Public a functioning component.

    Consider the parallel from the previous post. AI operating in governance without compliance with the Seven ROL Compliance Categories produces a coordination defect — the instrument cannot carry valid legal foundation to the next component in the chain. The vote faces the same mechanism. When the instruments governing the vote fail against the Seven Categories, the Public cannot exercise its coordination function. Not because the vote was taken away. Because the information architecture that makes valid participation possible was not there.

    The vote is the Public’s most recognised instrument for exercising its coordination function. Its result is a statutory instrument — a proclamation, an instrument of appointment, a gazette notice of electoral outcome — that shifts the functional weight of the components of state relative to each other, where statute requires. That instrument must satisfy the Seven ROL Compliance Categories to produce a valid coordination effect. When it does not, the coordination effect is defective. Not stolen. Not suppressed. Defective. And that defect is measurable.

    The SROL’s measurement instrument does not measure whether the outcome of the vote was good. It measures whether the instruments governing the vote satisfied the observable conditions for valid coordination. An election conducted under instruments that failed against the Seven Categories carries a coordination defect regardless of who won — and regardless of how the count went. The measurement does not reach the count. It reaches the instruments. Systems can exist in prolonged disequilibrium — not collapsing, not recovering, just managing — and the SROL’s measurement instrument records that displacement without judging its direction. The measurement is neutral on outcome. That is what makes it a measurement rather than a political position.

    The same seven questions apply to every governance instrument in the Public’s coordination chain: Was it authorised? Was the authority within jurisdiction? Was it clear? Was the Public consulted? Was it published in accessible form? Did it correctly identify the instrument that gave it its power? Was it properly commenced? The Seven Categories do not ask whether the election was free and fair. They ask whether the instruments governing it satisfied the observable conditions for valid coordination. The answer is either yes or no — and it is in the documentary record.


    The Implement Already Exists

    The Seven ROL Compliance Categories are the implement — not a new tool to be built, not a clause to be negotiated between parties. The implement already exists — published open access, empirically grounded against 28,337 gazette instruments (Seven ROL Compliance Category methodology, Felgate R. (2026). DOI: 10.5281/zenodo.21134975), available to any member of the Public today.

    Public Participation is Category 4 of the Seven ROL Compliance Categories. Its failure is not a political grievance. It is an observable coordination defect — identifiable in the documentary record, verifiable against the gazette, producible as evidence by any person who can point to the instrument and demonstrate the failure.

    Three things become possible the moment the Public holds this implement:

    In court: a documented Category failure in a governance instrument — the electoral regulation that was never properly published, the campaign finance rules that were never consulted on, the commencement notice that was signed but never gazetted — is evidence of a coordination defect in the documentary record. The standing question — who may bring such evidence before a court, and on what basis — is one the ROL literature has not yet answered. But the implement is available before that question is resolved. The evidence is producible now.

    In consultation: the Seven Categories applied to a proposed regulation produce a verifiable submission — not opinion, not political preference, but a documented coordination assessment. A Public that can verify whether a proposed regulation satisfies the Seven Categories before it is gazetted participates in governance at the point of production, not after the fact.

    In the electoral process: the Seven Categories applied to the instruments governing the electoral process produce a verifiable record of whether those instruments satisfied the participation condition before the vote occurred. Not after the count. Before. That is a qualitatively different kind of Public participation — prospective, documentary, verifiable.

    The Public is not the audience for this argument. It is a fourth component of state — with a defined coordination function, a published measurement standard, and an implement already in its hands.


    The Public Through AI

    The Public cannot apply the Seven Categories to every governance instrument alone. 28,337 gazette records represent sixteen years of systematic documentary measurement. The Public — exercising its coordination function, with AI as its instrument — can apply the implement at scale.

    The verification programme described in the previous post — which applies the Seven Categories automatically and returns a verdict on each — can make Category compliance verifiable for the Public across the full documentary record, in the same way it makes the ROL due diligence clause self-enforcing for AI suppliers in government procurement. Input: a governance instrument. Output: pass or fail on each of the Seven Categories, overall verdict, identified coordination defects. Not AI deciding what the Public should know. The Public applying the Seven Categories through AI — AI as the instrument through which that agency operates, not as the agent.

    The Public’s institutional architecture is public space — the open, distributed, accessible space within which its coordination function is exercised. Public space, public record, public discourse, public proceedings: these are the institutional architecture within which any person can exercise the Public’s coordination function, with AI as that person’s instrument. That architecture is not concentrated in a building or an institution. It is diffuse and open — accessible to any member of the Public, not by appointment or credential, but by the observable conditions that define valid participation. The individual falls within that architecture — and so does AI, as the instrument through which that individual exercises the Public’s coordination function at scale. That is its strength and its vulnerability: open enough for any person to exercise the Public’s function, open enough for any actor to degrade the conditions that make that exercise valid.

    The vote is the most recognised of the Public’s coordination instruments. It is not the only one. Every governance instrument that fails any one of the Seven Categories — every regulation, notice, and proclamation produced without the Public’s required participation, without proper authority, without clear jurisdiction, without adequate publication — is a site where the Public’s coordination function has been denied. The implement identifies them. AI makes the identification practical at scale.

    The rights and freedoms through which the Public participates in governance — the right to vote, to assemble, to access information, to participate — are not entitlements created by constitutions. They are the Public’s coordination function list, recognised by constitutions. The thick rule of law individualised those functions into personal claims. The SROL’s four-component model restores them to what they always were: the coordinate governance functions of a fourth component of state. Those functions have always existed. The implement to measure whether the state satisfies them now exists too. The SROL’s measurement instrument makes them visible — and actionable.

    The functions are old. Their formal recognition is not. The Public as a formal participant in governance — not just at election day but in every instrument the state produces — is a genuinely new ontological position. Its institutional expression is still emerging. AI may be the first form that expression takes.

    The Rule of law describes a closed system of governance. Its four components are components of state — not arms of state, not components of governance per se — each with distinct functions and separate powers that are together both necessary and sufficient to governance. The Public is one of those four. Not the audience. Not the residual. A component — with functions, with powers, with an implement, and with an instrument to use it at scale.


    The Science-based Rule of law framework has been in development for twenty years. Its systematic documentary measurement covers sixteen years of South African government gazette data, applying formal scientific methodology to governance. The Seven ROL Compliance Category methodology is published open access: Felgate, R. (2026). DOI: 10.5281/zenodo.21134975.

    Rita V. Felgate is an independent legal practitioner and governance researcher. She is the founder of ruleoflaw.science and the developer of the Science-based Rule of law framework.

  • The Contractual Standard That Would Make AI Governance Workable

    ROL: Use it or lose it

    Rita Felgate | August 2026
    ruleoflaw.science

    An em-dash publication: developed through human-AI cognitive synthesis with Claude (Anthropic). Learn about the em-dash methodology


    AI is already operating inside the information architecture of states. What has to be resolved urgently is what governs it when it does.

    States based on the separation of powers depend on their information architecture to coordinate access to and exercise of power. Over centuries, key attacks against states have targeted that architecture. In antiquity it was by smashing clay tablets bearing documentary evidence of state laws, administrative records and decisions — the Western Roman Empire’s governance failure was preceded not by conquest alone but by the breakdown of the documentary record that held its coordination together. History is littered with states and empires that empire building and information destruction leave behind — their coordination collapsed before their armies did.

    Attacks on state information architecture run from obvious to covert — espionage, propaganda, and the flooding of decision-making with false or distorted intelligence. Those attacks can emanate from other states or non-state actors.

    What AI introduces is not a new threat to state information architecture, but an old threat that can be implemented at new scale and speed.

    Several European governments and agencies have ended or restricted their use of Palantir systems — France’s domestic intelligence agency, Germany’s military cyber command, Switzerland (which repeatedly declined Palantir contracts on sovereignty grounds), and others — following concerns about data sovereignty, single-supplier dependence, and the terms on which state coordination information was being processed. Governance experts have raised significant concerns about the documentary basis on which DOGE operated within federal agencies. Both cases point to the same structural problem: AI operating in government without the documentary record that makes its exercise of state power verifiable. That is informational conquest — capture or disruption of a state’s information architecture without physical destruction, through means that are deniable, incremental, and legally ambiguous. It is the nuclear option of the information age: without crossing a border, without firing a shot, and without triggering the international legal responses that physical conquest would attract.

    AI suppliers entering the coordination chain face a consequence they may not have anticipated: their systems are not merely tools deployed by the state, but instruments through which the state accesses and exercises power. Palantir’s CEO, Alex Karp, told investors in February 2025 that the company exists to improve the institutions it partners with and, when necessary, “to scare our enemies and, on occasion, kill them.” That is not a commercial function. It is an exercise of state power — and it must satisfy the same observable conditions as any other exercise of state power, including the highest-stakes ones.

    However, while the information age creates the vulnerability, it also facilitates a solution. A defective governance coordination instrument can be identified at the point of its production. States do not have to wait for information system collapse before beginning reconstruction of their information architecture — instrument failures can be identified and repaired immediately, before they accumulate past the point of correction.

    The Rule of law is a system of governance that provides the coordination architecture and informational integrity upon which state stability depends. It is not a binary presence or absence, but a system of governance whose parameters are measurable, whose failures are correctable, and whose integrity is maintainable in real time. It has remained unnamed, unmeasured, and undefended — until now. A commencement failure can be measured across sixteen years of gazette records; an authority defect can be corrected by republication. That is what measurement makes possible. The previous post established how that measurement works and what it has found.

    The introduction of AI within government operations is currently negotiated through procurement contracts. With the vulnerability so visible, two questions are being asked simultaneously about those contracts.

    Governments are asking: how do we ensure AI is governed through legislation — and faced with a threat they cannot meet by force, how do they protect the information architecture on which the state depends? AI suppliers are asking: how do we manage our exposure of deploying in government?

    The fundamental answer to both questions lies not in new legislation or new regulatory bodies but in understanding and applying the ROL as the coordination rule that has been governing and stabilising states all along. The ROL is the coordination rule whose conditions are observably satisfied or not satisfied in the documentary record — with the particular advantage that both parties can independently assess whether those conditions are met, without requiring trust, common interest, or negotiated agreement. Its conditions are necessary to the coordination capacity of the state and hence its stability. Both parties to a procurement contract enter this era vulnerable to what has just been described. Both have an equal interest in the ROL framing a solution.

    A note on terminology: throughout this post, “Rule of law” does not mean democratic values, judicial independence, or human rights — though these may be its effects. It means the coordination mechanism by which access to and exercise of state power is measured, prior to and independently of any particular piece of legislation. If that distinction is unfamiliar, the previous post sets it out.


    The ROL Is Upstream of the Contract

    The ROL arises with the complexity of governance systems that makes the separation of powers between components of state inevitable. It is what enables that complexity to maintain stability. It preceded the legislation that expresses it — and it precedes any contract that rests on those.

    The ROL exists upstream of and a priori to legislation and state contracts. Their coordination effect is conditional on the ROL’s observable conditions. Those conditions are not conditional on them. Legislation or state contracts that ignore, contradict or misconstrue the ROL’s observable coordination conditions have outputs that cannot enter ROL-compliant information streams — they fail the coordination test that valid governance instruments must pass.

    When a government and AI supplier contract to provide government services, both parties turn to legislation and contract law for the framework of their obligations, and their due diligence is perceived as limited to those instruments. But the ROL sits upstream of both.

    An AI supplier contracting with government to provide government services should therefore assess compliance at two levels: firstly, whether the legislation underlying the contract satisfies the ROL’s observable coordination conditions, and secondly, whether the contract itself, in what it authorises the parties to do, is consistent with those conditions. Likewise, a government official contracting with an AI supplier needs to ensure the same.

    A contract signed by an official without valid statutory authority is not saved by its black letter terms. A contract authorising AI to carry out functions that exceed the statutory powers of the contracting state entity is not saved by the fact that both parties agreed to it. The AI supplier enters the coordination chain at the point its output is used in the state’s access to or exercise of power — not merely because it supplies AI. At that point, the observable conditions for ROL-compliant coordination apply to its output.


    Gap Between Constitutions and the Rule of law

    The reason neither governments nor AI suppliers have yet invoked a ROL clause is simple: the ROL has never been operationally defined in a form that a contract could reference. Some Constitutions invoke it — South Africa’s Constitution names it a foundational value, full stop, no definition — and courts have developed the concept through recognition in judicial precedent. The SROL’s Seven ROL Compliance Categories are the first operationalisation precise enough to be contractually referenced. That changes what is now possible.


    The Observable Conditions for ROL-Compliant Coordination

    Every documentary governance instrument — every Act of Parliament, regulation, ministerial notice, proclamation, commencement notice, ministerial direction, assignment, delegation, and notice of general application — must satisfy seven observable conditions for ROL-compliant coordination to be possible. That is not a finding; it is what coordination requires. Without any one of these conditions, access to and the exercise of state powers cannot coordinate a ROL-compliant effect. The Seven ROL Compliance Categories: Authority, Jurisdiction, Clarity, Public Participation, Publication, Referent, Commencement. Full definitions and codebook →

    Many of these conditions are formally codified in legislation — but their source is the ROL, not the legislation.

    What research establishes is what happens when the conditions fail.

    The failure of any Category means the instrument cannot coordinate a ROL-compliant effect. The question of whether domestic law treats the resulting instrument as void, voidable or otherwise remediable is a separate remedial question answered by positive law. The coordination failure is not a separate question — it is the observable fact that the instrument cannot do what it is supposed to do.

    Twenty years of research. 28,337 South African government gazette documentary records coded against the Seven ROL Compliance Categories. A correlation of −0.979 between the integrity of the information architecture those categories measure and the breakdown of governance coordination. That integrity measure is built from three of the seven categories — Publication, Referent and Clarity. The independence test shows their relationship with coordination failure is significantly stronger than that of the remaining four.

    This research demonstrates that the Rule of law coordination mechanism is not theoretical. It is measurable. It has been measured.


    Risks the AI Era Is Exposing

    AI is now generating governance instruments — drafting regulations, processing ministerial notices, producing policy documents that become the basis for the exercise of state power. AI is also making decisions that are properly those of creatures of statute — decisions requiring statutory authority, affecting rights and obligations, depending on powers conferred by legislation, and which carry the duties and obligations of creatures of statute in performing those decisions. Whether generating instruments or making decisions, the ROL’s observable coordination conditions apply.

    Much of what applies does so already through existing black letter law and common law. What is not yet required by contract with governments is systematic verification of Seven ROL Compliance Category compliance at the point of AI production.

    This absence creates exposure on both sides of the contract.

    For the state: a governance instrument generated or processed by AI that fails any one or more of the Seven ROL Compliance Categories means that state powers exercised down the line depending on that instrument cannot coordinate a ROL-compliant effect. A failure to commence an Act, for example, cannot be the basis of and cannot coordinate any exercise of power under that Act. To the extent that any exercise of state power depends on such an instrument as its legal authority, that exercise is without a secure legal foundation and remains challengeable.

    For the AI supplier: the supplier enters the coordination chain when its output is used in the state’s access to or exercise of power. From that point, the observable conditions for ROL-compliant coordination apply. Where conditions of AI deployment or its product fail any one or more of the Seven Categories, the contract terms that give that product its statutory effect leave the AI supplier without a secure legal foundation.

    What flows from these failures — whether invalidity, damages, rescission, judicial review, administrative correction, or nothing at all — is answered by positive law in each jurisdiction, not by the ROL. The ROL establishes the coordination failure. Positive law determines the consequence.

    The insertion of a ROL due diligence clause within procurement contracts between AI suppliers and government addresses a different and prior question: it creates an express contractual duty to test and document compliance with the Seven Categories, and allocates consequences between the parties if that duty is not performed. That clause — and what it makes possible — is the subject of the next section.


    ROL: A Double-Edged Due Diligence Sword

    The ROL’s coordination conditions operate in both directions simultaneously — and this is the distinctive strength of ROL-based requirements within a procurement contractual instrument.

    Satisfying the observable conditions for ROL-compliant coordination is inherent in the statutory function. The state cannot choose whether to require that compliance when AI is delivering functions the state provides under statutory appointment. The ROL due diligence clause — the contractual instrument of the state — simply makes that pre-existing condition explicit, documented, and with contractually allocated consequences for failure.

    The AI supplier’s accountability to the ROL’s coordination conditions exists a priori the government procurement contract — neither created by the contract nor imported into it, but simply expressed by it. AI output in a governance function occupies the position of witness evidence: it is evidence of the state’s exercise of power and must satisfy the same Seven ROL Compliance Category conditions as any other governance instrument through which state power is coordinated.

    A state that refuses to include the ROL due diligence clause is not taking a legal position. It is accepting instruments into its documentary record without verification that they satisfy the observable conditions for ROL-compliant coordination.

    An AI supplier that resists the clause is not protecting a commercial position. It is declining to document that its exercise of government functions or its product satisfies the conditions that coordination requires.

    Neither party’s position is defensible when a coordination failure becomes visible — in a courtroom, a regulatory proceeding, or a constitutional challenge. The resolution of coordination failures runs through the ordinary functioning of the components of state — the Public, the Legislature, the Executive, and the Judiciary. That is where the standing question will be answered: in positive law, not in the ROL.

    The standing question becomes urgent immediately precisely because ROL coordination failures accumulate undetected and unresolved: they pile up daily — in uncommenced Acts, in unpublished regulations, in instruments that lack their authority — and go unrecognised until someone with a specific legal grievance stumbles into court a decade later, if they surface at all. In circumstances of state capture, they may never surface because the components of state that should bring them have themselves been captured.

    The SROL’s ROL measurement instrument changes what can be put before a court. A documented Category failure is evidence — observable, verifiable, grounded in the documentary record — of a defect in the information architecture on which the exercise of separate powers of state can coordinate. That is not a private grievance. It is a failure in the coordination on which separately exercised state powers depend. Whether the ordinary requirements of standing fit that kind of evidence — where the defect is public and documentary rather than private and consequential — is a question for positive law. It is one the ROL literature has not yet answered.


    Nature of the ROL Due Diligence Clause

    The clause does not make the ROL enforceable. It makes ROL due diligence contractually enforceable.

    As a contractual instrument, the ROL due diligence clause makes the a priori ROL accountability explicit and documented without requiring new legislation, a new regulatory body, or international agreement.

    It requires one ROL due diligence clause in an AI procurement contract, referencing the published Seven ROL Compliance Category methodology (Felgate, R. (2026). DOI: 10.5281/zenodo.21134975), specifying that AI-generated or AI-processed governance instruments must satisfy the Seven ROL Compliance Categories during due process, and providing for independent verification against the documentary record.

    The Seven Categories are not equally within either party’s control — and that differential allocation enhances the clause’s discriminating power, not its limitation.

    The AI supplier’s production-side categories — Clarity, Publication, Referent — are within the AI supplier’s control at the point of instrument production. The AI supplier warrants these. The process-side categories — Authority, Jurisdiction, Commencement — are properties of the state’s upstream process: whether the official held the power, whether the instrument fell within it, whether the commencement step was taken. The state warrants these. Public Participation concerns a process that preceded the instrument’s production entirely; both parties verify against the documentary record that participation occurred — the record of it, which is what a contract can reach. Where participation was conducted by a third party acting under statutory authority, that party’s outputs are themselves instruments of the state’s exercise of power and must satisfy the Seven Categories in their own right.

    Where either party cannot verify a Category, that inability is itself the signal of due diligence failure. If the state cannot verify its own Authority or Jurisdiction, the verification gap is the defect. If the supplier cannot verify Commencement, the uncommenced status is the defect. The verification failure and the Category failure are the same observable fact, seen from different positions in the coordination chain.

    Verification against the published codebook and the state’s gazette documentary record is possible today — manually, by any person with access to both. The verification programme makes verification continuous and automated. The practical path to auditing due diligence will likely be by third-party verification: an independent party applying the published codebook to the state’s gazette documentary record, with no stake in the outcome and consequences available against either party. That is the ROL due diligence clause in its first operational form — and it is the market-setting move.

    Indicative clause elements. A ROL due diligence clause should specify: (1) that AI-generated or AI-processed governance instruments are subject to Seven ROL Compliance Category verification before entry into the state’s documentary record; (2) which Categories each party warrants (production-side: supplier; process-side: state; Public Participation: verified jointly against the documentary record); (3) that where verification is impossible, the gap constitutes a Category defect triggering the clause’s consequence provisions; and (4) that verification may be performed by an independent third party against the published codebook (Felgate, R. (2026). DOI: 10.5281/zenodo.21134975) and the gazette record.

    For AI suppliers, the clause is not a burden, but a documented due diligence trail that converts a coordination failure from a liability into a defence.

    The first government that includes such a clause sets a precedent. But the more likely first mover is the supplier. An AI supplier operating across jurisdictions has a direct commercial interest in one documented standard rather than twenty bespoke contractual regimes. Refusing to document ROL compliance is indefensible when a coordination failure surfaces. Accepting the standard — and being able to demonstrate compliance against a published, open-access codebook — is the competitive differentiator in government procurement.

    The window is open. Governments are actively seeking AI governance frameworks. The ROL’s observable coordination conditions are published, open access, empirically grounded, and contractually operationalisable today.

    The clause does not require immediate full compliance. It requires documented due diligence — a contractually visible trajectory that both parties can monitor, enforce, and improve against. And it does not require the state to solve its compliance deficit before contracting: it creates the commercial space for AI suppliers to bring production-side compliance tooling with them — the forms and pre-publication checklists that make Seven Category compliance achievable at the point of production. Verification instruments, however, must remain independent — the party being verified cannot supply the verification standard.


    The Verification Programme

    The ROL due diligence clause is the immediate step. The programme is the destination.

    What is in store if this fails is not a new dystopia. It is the acceleration of the existing one — more winner-takes-all transactionalism, faster consolidation of power by those who control the information architecture, less time for correction before the degradation becomes irreversible. We are on the precipice already — a moment where states can choose a rules-based order or face the real risk of information system failure and the discoordination of state functions that follows. The direction is visible.

    A programme that applies the Seven ROL Compliance Categories automatically — to any governance instrument, during due process, before publication — is technically buildable from the published methodology. Input: a governance instrument. Output: pass or fail on each of the Seven Categories, overall verdict, identified coordination defects, suggested corrections.

    The question that concerns every state is whether governance AI will operate within the coordination architecture the state requires — or outside it. The SROL’s answer is structural: the observable conditions for ROL-compliant coordination apply to the instrument regardless of who or what produced it. The Public, the state and its service providers undertaking governance functions — as agents, with AI as their instrument — apply those conditions through the verification programme. That is not a constraint on what AI can do. It is the coordination architecture within which what AI does either serves governance or fails it.

    The ROL due diligence clause does not wait for legislation. It does not wait for a regulator. It does not wait for international agreement. It waits for two parties to agree. That route is available today. Use it.


    The Science-based Rule of law framework has been in development for twenty years. Its systematic documentary measurement covers sixteen years of South African government gazette data, applying formal scientific methodology to governance. The Seven ROL Compliance Category methodology is published open access: Felgate, R. (2026). DOI: 10.5281/zenodo.21134975.

    Rita V. Felgate is an independent legal practitioner and governance researcher. She is the founder of ruleoflaw.science and the developer of the Science-based Rule of law framework.

  • Unnamed, Unmeasured, Undefended: The Rule of Law in Plain Sight

    An em-dash publication: developed through human-AI cognitive synthesis with Claude (Anthropic). Learn about the em-dash methodology →


    Editor’s note: This post was updated on 29 July 2026 to refine several claims following peer review. The substantive argument is unchanged.

    A governance instrument arrives. It may be an Act of Parliament, a ministerial notice, a proclamation, a regulation. It faces seven immediate questions:

    Does it identify the authority under which it was made? Does it fall within the jurisdiction of that authority? Is it clear enough to act on? Was the public given opportunity to participate in its making? Was it published in a form that reaches those subject to it? Does it correctly identify what it amends or replaces? When does it take effect?

    Where a category is applicable, its absence establishes a coordination defect at the point of production. Whether positive law treats the resulting instrument as void, voidable, suspended or otherwise remediable is a separate remedial question. What is not separate is the defect: the coordination the instrument was supposed to enable — between the exercise of power and its effect — has no secure legal foundation and either does not happen, or happens without a legal basis and remains permanently challengeable.

    This is not only a theory. It produces a measurement. Twenty years of research, including sixteen years of systematic documentary measurement applied to 28,337 South African government gazette documentary records, produced a correlation of −0.979 between the integrity of that information architecture and the breakdown of governance coordination.

    The river is predictable from the source.


    The Wrong Donkey

    Two demands are sounding louder in every jurisdiction and across every international forum.

    The first: where is the Rule of law? Governance is failing, institutions are captured, laws mean nothing in practice, the powerful act without consequence.

    The second: we need a rules-based international order. Might is prevailing over right, multilateral institutions are paralysed, states transgress each other’s sovereignty and independence without sanction.

    Both demands are real. Both are urgent. And both are pinning the tail on the wrong donkey.

    Democracy has been tried, tested, and has not delivered stable, coordinated, contestable governance at the scale the world now requires. It has been blamed as author of its own failures.

    But it is not the Rule of law that has failed — it is its understanding. It has also been blamed for failing to deliver the Rule of law. It has been governing stable states all along, and without understanding it has remained unnamed, unmeasured, and therefore undefended.

    “The Rule of law has not failed. Its understanding has.”


    Descent of the ROL Model into Teleology and Tautology

    A dominant post-Cold-War reform model treated democratisation and Rule-of-law institution-building as causally reinforcing: establish the Rule of law and democratic governance follows. Post-Soviet experience exposed the inadequacy of that model — formal institutions did not reliably produce either. The response was not to abandon the claim but to reverse it: democratise first and the Rule of law follows. That also failed.

    From there, the descent of the model into teleology and tautology was inevitable. First, enter a teleological ghost that recast the ROL and democracy as expressions of higher ideals — human freedom, dignity, flourishing — that exist outside the observable record. The model became untestable and unverifiable. The direction of cause and effect, in either direction, left undifferentiated and undifferentiatable. Enter tautology: democracy and the ROL became indistinguishable — two sides of the same coin.

    The current impasse: democracy defines the Rule of law; the Rule of law defines democracy. Neither is independently testable. The model can persist indefinitely — and it has.

    “The model can persist indefinitely — and it has.”

    What the world inherited is what that claim left behind when it failed.


    What the Rule of Law Actually Is

    The Rule of law is not democracy. It is not a set of values or aspirational principles. It is the coordination mechanism that arises from the separation of powers within a state — it is the rule that access to and exercise of powers must be based on reason, rationality and logic (RRL) necessary to the equilibrium and continuity of a state.

    The coordination requirement it embodies is logically prior to any particular constitutional or democratic embodiment of it. It is what makes coordination between components of state exercising separate powers possible. Without it, no political form can deliver stable, predictable, contestable governance.

    And it is measurable. Not through indices that aggregate impressions of governance quality against democracy. But directly through the state’s documentary record — the gazette instruments and judicial precedent through which state power actually operates. Seven categories. Yes or no. At the moment of production.

    The South African National Energy Act 34 of 2008 contained a commencement provision. Section 6 — the operational heart of the Act — required a proclamation before it could take effect. That proclamation was not issued. For sixteen years, energy planning proceeded on a legal basis that did not exist. For sixteen years, judicial judgements relating to the Act were decided as if it had a legal basis.

    The a priori question is not: what caused South Africa’s energy crisis? That is a complex systems question. The a priori question is: was the commencement proclamation published? The answer is no. The consequence is direct — to the extent that any exercise of power depended on Section 6 as its legal authority during those sixteen years, it was legally unauthorised. The correction was equally direct: in 2024 the proclamation was finally issued and Section 6 took effect. Sixteen years after enactment. The cause was in the document. The correction was in the document.


    The Same Answer to the Two Loud Demands, at Two Scales

    The Rule of law as instrument answers both demands simultaneously — not because it was designed to, but because both demands are asking the same question from different points of view.

    The intrastatal question: are the governance instruments operating on this population based on reason, rationality and logic? Are they published in forms the population can find, read, and act on?

    The interstatal question: does this state’s exercise of power stay within the limits that the Rule of law defines for it? A state whose governance instruments consistently fail the seven categories is a state that exports instability. If anything, the opposite is true of states that comply: low Rule of law compliance is not just a domestic governance problem. The SROL predicts that it is the mechanism by which extraterritorial threat is generated — a prediction that cross-state testing will either establish or falsify.

    “Low Rule of law compliance is not just a domestic problem. The SROL predicts it is the mechanism by which extraterritorial threat is generated.”

    Same standard. Same instrument. Two scales.


    A Rules-Based Order That Does Not Depend on Power

    The Rule of law — independently defined, measurable, observable in the gazetted documentary record — provides a standard for assessing state conduct that does not depend on the relative power, alliance, or democratic credentials of the states concerned. That standard is the basis for a rules-based order.

    A state’s gazetted documentary record is a real-time measure of its compliance. Compliance with the Rule of law is a measure of a state’s stability. Stability is a measure of the risk a state represents to the domain and authority, and hence the sovereignty and independence, of every other state.

    States that strive to meet the ROL documentary standard constitute a Commons — a space of governance that does not depend on the goodwill of the powerful, on winner-takes-all transactional alliances, or the aspirations of any ideal outside the observable record.

    The Rule of law has been made invisible by teleological and tautological models that pin it to the wrong donkey. The Rule of law is not missing. It is misunderstood.

    It is now measurable. That changes everything.


    The Science-based Rule of law framework has been in development for twenty years. Its systematic documentary measurement covers sixteen years of South African government gazette data, applying formal scientific methodology to governance. The empirical foundation is published open access: Felgate, R. (2026). DOI: 10.5281/zenodo.21134975.

    Rita V. Felgate is an independent legal practitioner and governance researcher. She is the founder of ruleoflaw.science and the developer of the Science-based Rule of law framework.

  • The Rule of Law is the Next Big Thing

    The ROL: Talking Truth to Power
    Rita Felgate | June 2026
    ruleoflaw.science

    An em-dash publication: developed through human-AI cognitive synthesis.


    Every generation has its governance breakthrough. The one that was going to fix the problem of power — make it accountable, constrain it, subject it to something larger than itself.

    Constitutionalism was going to do it. Human rights law was going to do it. Democratisation was going to do it. AI regulation is the current candidate.

    Each framework devised arrived with genuine intellectual force. Each hit the same wall.

    The wall is always the same wall: undefined terms, no operational standard, no verification, no enforcement. “Democratic values.” “Constitutional principles.” “Rule-based governance.” “Responsible AI.” Everyone endorses them. Nobody can operationalise them. When it matters — when power is actually being exercised over real people in real contexts — the terms dissolve into competing interpretations, each as defensible as the next, none of them resolvable by the framework that was supposed to resolve them.

    The Rule of Law is different. Not because it is newer — it is ancient. But because it is, for the first time, measurable.


    (more…)
  • The AI Moratorium Problem: Why the Rule of Law Answer Couldn’t Be Named

    The ROL: Talking Truth to Power
    Rita Felgate | June 2026
    ruleoflaw.science

    An em-dash publication: developed through human-AI cognitive synthesis.


    A. The Crisis Is Already Live in Governance

    The AI governance crisis is not waiting in the future. It is already live inside the machinery of state.

    AI systems are already being used to process benefits, generate enforcement recommendations, assist border-control decisions, support policing and intelligence functions, and operate inside public-sector information systems. These are not private conveniences. When AI performs or supports a public function under state authority, it participates in the exercise of state power.

    (more…)