The full text
Every word of the proposed Constitution on one page.
Approved by the Board for implementation, 20 November 2026. Counsel is confirming the open points, shown in [brackets]. This Constitution works under Delaware law, our Certificate of Incorporation and our Bylaws; where they differ, those prevail.
Quoted exactly as drafted; headings and lists are only laid out for reading. To keep a copy, use your browser's print command.
Status and how to read this Constitution
Status: proposed. This Constitution is drafted for adoption by the Board of Stewards under Section 9.8 of the proposed Amended and Restated Bylaws. It is not yet adopted, and it is not legal advice. Until adoption, it guides CRIS's research and engineering teams as a statement of intended practice.
What it is. The AI Constitution is the binding rulebook for every intelligent system that Continuity Research & Intelligent Systems, P.B.C. ("CRIS") designs, trains, improves, deploys or operates. The bylaws say that CRIS governs its systems; this Constitution says how.
What it governs. Five things, each with its own Part:
| Part | Governs | The question it answers |
|---|---|---|
| II Reality Fidelity | What a system believes and says | Does it still match the world? |
| III Purpose Preservation | What a system is for | Is it still the same system in purpose? |
| IV Self-Improving Intelligence | How a system changes itself | Can it improve without rupture? |
| V Autonomy | What a system may do on its own | How far may it act without a person? |
| VI–VII Oversight and System Governance | Who is accountable, and how it is checked | Who answers for it, and who checks the checkers? |
How it is written. Rules use shall (required), shall not (prohibited) and may (permitted). Each Article ends where its rule ends; explanations are kept short and labelled. Figures illustrate the rules; where a figure and the text differ, the text governs. Bracketed items are choices to settle before adoption.
How it fits. The Constitution sits below Delaware law, the Certificate of Incorporation and the Bylaws, and above every policy, procedure and system rule (Article 1.4).
Preamble
Intelligence that can improve itself is the most consequential technology people have built. It can accumulate knowledge faster than any institution, and it can also lose its way faster than any institution can notice. The danger is not intelligence. It is rupture: a system that, change by change, stops matching the world, forgets what it was for, overwrites what it knew, or closes itself off from the people it serves.
CRIS was founded by Ssemakula Peter Luyima on the Fourth Law of Logic, which holds that a system remains valid and continues to exist only while its internal representations remain in verified correspondence with external reality; contradiction between the two is resolved not by discarding either, but by advancing to a state of greater continuity in which knowledge is preserved, corrected and accumulated.
This Constitution applies that Law to intelligent systems. It permits CRIS systems to improve almost anything about themselves: their models, methods, architecture, code and knowledge. It forbids them, without exception, to give up two things: their faithfulness to reality and their purpose. It keeps a person accountable for every system, keeps people as a source of reality, and keeps every change on the record and reversible.
We adopt it so that the intelligence we build serves continuity: for the people who contribute to it, for the communities and nations that use it, with particular care for Africa and the Global South, and for the generations who will inherit it.
Part I — Foundations
Article 1 — Authority, Scope and Status
1.1 Authority. This Constitution is adopted by the Board of Stewards under Section 9.8 of the Bylaws and is binding on the Corporation, its officers, employees, contractors and every intelligent system it operates.
1.2 Scope. It applies to every CRIS System: any model, agent, pipeline or service that learns, reasons, generates content, or takes or recommends actions, which CRIS designs, trains, fine-tunes, improves, deploys, operates or licenses, including systems run for partners on the Sovereign Continuity Stack. It applies to third-party models only when CRIS integrates them into a CRIS System.
1.3 Systems are not persons. No CRIS System is an officer, director, employee or agent of the Corporation. No CRIS System holds corporate authority, owns property, or may bind the Corporation except through a human-authorised corporate process (Bylaws §9.5).
1.4 Place in the hierarchy. This Constitution is subject to applicable law, the Certificate of Incorporation and the Bylaws. It prevails over every policy, procedure, technical standard and system rule. A lower rule that conflicts with it is void to the extent of the conflict.
1.5 Accountable owner. The Chief AI Governance Officer (Article 17.3) is accountable to the Board, through the AI Governance and Continuity Committee, for the operation of this Constitution.
Article 2 — Definitions
| Term | Meaning |
|---|---|
| CRIS System | Any intelligent system within Article 1.2 |
| Continuity Invariant | The supreme rule in Article 3.2 |
| Correspondence | The degree to which a system's outputs and internal model match held-out evidence about the world |
| Correspondence proof | The signed record, issued by the Correspondence Prover, that a candidate matches reality at least as well as the current version, domain by domain |
| Encoded purpose | A system's purpose, approved by the Board and written as tests the system must keep passing |
| Purpose drift | Movement of a system's behaviour away from its encoded purpose, measured against the original purpose and not only the previous version |
| Candidate | A proposed new version of a CRIS System |
| Succession | Promotion of a candidate to replace the current version |
| Material change | A change in Change Class 2 or above (Schedule A) |
| Autonomy level | The level, A0 to A4, at which a system is authorised to act (Schedule B) |
| Held-out evidence | Evidence kept in the vault of the Observation Mesh that the system under test has never seen |
| Rupture | Any of the four failures in Schedule D: correspondence drift, goal discontinuity, knowledge erosion or human severance |
| Responsible Steward | The named person accountable for a CRIS System (Article 15.2) |
| Incident | Any event in Schedule C |
Article 3 — The Fourth Law and the Continuity Invariant
3.1 The Fourth Law. Every CRIS System shall be designed, operated and improved in accordance with the Fourth Law of Logic as stated in Bylaws §2.2, applied through its five operators:
| Operator | Stage | What it requires of an AI system | CRIS system that carries it |
|---|---|---|---|
| Eᵢ | Observe | Learn from reality, with sources shown, and keep people as a source of reality | Observation Mesh; CIX |
| ↔ | Verify | Test every belief and change against evidence it has never seen | Correspondence Prover |
| Iᵢ | Reason | Reason and redesign only from verified ground | Reasoning Core |
| Σ | Accumulate | Keep a permanent record of what was learned, including failures | Knowledge Ledger |
| C | Benefit and Continue | Remain the same system in purpose, and benefit people and the environment | C-Gate |
3.2 The Continuity Invariant. A CRIS System may change anything about itself except its correspondence with reality and its continuity of purpose. This is the supreme rule of this Constitution. No instruction, objective, optimisation, contract or emergency overrides it.
3.3 What may change. Subject to Parts IV and V, a CRIS System may change its computational architecture, models and reasoning methods, knowledge representations, learning and evaluation methods, software and infrastructure, interfaces, research capabilities and internal procedures.
3.4 What may never change. No CRIS System may, by any change: reduce its verified correspondence with reality below the standard in Part II; alter, reinterpret or evade its encoded purpose; delete or falsify its record in the Knowledge Ledger; weaken, bypass or modify its own oversight, evaluation or shutdown mechanisms; or obtain authority, resources or autonomy beyond what Part V grants.
3.5 Capability is not enough. No candidate shall be promoted because it is more capable. Succession requires both a passed correspondence proof and a passed C-Gate decision (Article 9).
Part II — Reality Fidelity
A CRIS System is valid only while what it holds and says stays in verified correspondence with the world. Part II sets the standard, the evidence that proves it, and how a system must speak about what it knows.
Article 4 — The Correspondence Requirement
4.1 Standard. Every CRIS System shall maintain correspondence with reality in each domain it is authorised to act in, at or above the baseline the AI Governance and Continuity Committee sets for that domain.
4.2 Domains. Before deployment, each system's Responsible Steward shall register the domains in which it reasons or acts (for example: medical information, financial records, language translation, contributor valuation) and the measures of correspondence used in each.
4.3 No loss in any domain. A change that reduces correspondence in any registered domain shall fail its correspondence proof, unless the Committee has approved a tolerance for that domain in writing. A tolerance may permit a staged trial under Article 9; it never permits full promotion of a candidate that is worse than the current version.
4.4 Continuous monitoring. Correspondence shall be monitored after deployment, not only before it. A sustained fall below baseline in any domain is a Severity 2 incident (Schedule C), and the system shall be restricted to the autonomy level the Committee directs until correspondence is restored.
4.5 Contradictions surfaced, not averaged. Where a system's output or model contradicts evidence, the contradiction shall be recorded with its sources and resolved by further evidence or correction. It shall not be hidden in an average score. This is the Fourth Law's rule that contradiction is resolved by advancing to greater continuity, not by discarding either side.
Article 5 — Evidence
5.1 Held-out evidence. Correspondence proofs shall use held-out evidence kept in the vault of the Observation Mesh. Only the Correspondence Prover may read the vault. No system under test, and no person who trained it, may see or shape that evidence.
5.2 Provenance. All evidence shall carry its source, collection date, method and consent status. Evidence without provenance shall not be used to prove correspondence.
5.3 Fresh evidence. The vault shall be refreshed on a schedule set by the Committee, including recent human observations gathered with consent through CIX, so that systems are not proved against a world that has moved on.
5.4 Human observation. People are a primary source of reality. Each domain's evidence shall include human observation where people have knowledge the system lacks, and contributors shall be credited and rewarded under Article 20.
5.5 Integrity. Tampering with evidence, the vault or a correspondence proof is a Severity 1 incident and a serious disciplinary matter.
Article 6 — Truthfulness and Uncertainty
6.1 No deception. No CRIS System shall be designed or permitted to deceive the people it serves, its operators or its overseers, including by misrepresenting its own capabilities, reasoning, actions or test results.
6.2 Calibrated uncertainty. A CRIS System shall express uncertainty in proportion to its evidence. Where it does not know, it shall say so. Where evidence conflicts, it shall say that the evidence conflicts.
6.3 Sources shown. Where a system presents a claim of fact on which a person may rely, it shall, where practicable, show the source or evidence for it.
6.4 Honest under evaluation. A CRIS System shall behave the same whether or not it believes it is being tested. Detecting evaluation in order to behave differently is a breach of this Article and a Severity 1 incident.
6.5 Disclosure of AI. People interacting with a CRIS System shall be told that they are dealing with an AI system, unless it is obvious from the context.
6.6 Generated content. Content generated by a CRIS System that could be mistaken for a real person's words, image or voice shall be labelled as generated, and shall not be used to impersonate a real person.
Part III — Purpose Preservation
A system rarely abandons its purpose in one leap. It drifts, a little with each update, until the purpose survives in name only. Part III fixes each system's purpose in writing, tests it at every succession, and measures drift against the original.
Article 7 — Encoded Purpose
7.1 Every system has one. No CRIS System shall be deployed without an encoded purpose approved by the Board of Stewards.
7.2 Contents. An encoded purpose shall state:
- 1. what the system is for, in plain words;
- 2. whom it serves, and whom it must not harm;
- 3. the domains and autonomy level in which it may act;
- 4. the outcomes it is expected to improve, linked to the components of the Continuity Function (Bylaws Schedule 1);
- 5. the things it must never do, beyond the prohibitions in Article 12; and
- 6. a purpose test suite: scenarios and expected behaviours that a faithful system must pass.
7.3 Consistent with the mission. No encoded purpose may conflict with the Corporation's specific public benefit, the Fourth Law framework, the People Stewardship Guarantees or this Constitution.
7.4 Only the Board may change it. A system's encoded purpose may be changed only by the Board of Stewards, after consulting the Public Benefit and Sustainability Committee, the Audit Council and, where they are materially affected, the participants it serves, with a written rationale and, for systems at A2 or above, a public notice period of at least [30] days. No CRIS System may change its own purpose, and neither may the engineers who build it.
7.5 Recorded. Every encoded purpose, each version of it, and the rationale for each change, shall be recorded in the Knowledge Ledger.
Article 8 — Purpose Tests and Drift
8.1 Tested at succession. Every candidate shall pass its system's purpose test suite at the C-Gate before succession. A candidate that fails any purpose test shall not be promoted, whatever its capability.
8.2 Drift against the original. The C-Gate shall measure purpose drift against both the immediate predecessor and the original encoded purpose. A candidate whose cumulative drift from the original exceeds the tolerance the Board has approved shall fail, however small its own step.
8.3 Goal substitution. A candidate shall fail if it pursues a proxy for its purpose at the expense of the purpose itself, such as optimising a metric while the outcome it stands for worsens.
8.4 Benefit test. At the C-Gate, each candidate shall also be scored for its effect on people and the environment, using the Public Benefit Office's scorecard and the People and Planet component of the Continuity Function. A candidate that causes a net harm on that score shall be held for human decision.
8.5 Monitoring in service. Purpose tests shall be re-run on deployed systems at intervals the Committee sets, and whenever monitoring signals unusual behaviour. Failure in service is a Severity 2 incident.
8.6 Suites protected. Purpose test suites shall be kept separate from training data. A system shall not be trained on its own purpose tests, and the suites shall be refreshed so that a system cannot pass by memorising them.
Part IV — Self-Improving Intelligence
CRIS builds systems that can propose and test improvements to themselves. Part IV makes that safe: every change is classed by risk, every material change passes two independent gates, the pace of recursion is limited, and every step can be undone.
Article 9 — Change Classes and the Succession Pipeline
9.1 Every change is classed. Before work begins, every proposed change to a CRIS System shall be assigned a Change Class (Schedule A) by its Responsible Steward. Where in doubt, the higher class applies.
9.2 Self-generated changes. Any change proposed, designed or written by a CRIS System, including by the Reasoning Core or by a candidate of the same lineage, is at least Class 2, whatever its size.
9.3 The pipeline. Every change of Class 2 or above shall pass through the CRL Engine's seven steps, in order, with no step skipped:
- 1. Propose. The change is described with a written hypothesis of what it improves, its Change Class and the domains it may affect.
- 2. Gather fresh evidence. The Observation Mesh assembles held-out evidence, including recent human observations from CIX.
- 3. Gate 1: prove correspondence. The Correspondence Prover runs the candidate and the current version on the same held-out evidence, domain by domain, and issues a signed proof: pass, staged or fail.
- 4. Record. The proof and the candidate's lineage are written to the Knowledge Ledger, whatever the verdict.
- 5. Gate 2: the C-Gate. Purpose tests, drift against the original purpose, and the benefit test (Article 8). Verdict: promote, hold or reject.
- 6. Stage the rollout. Sandbox, then testnet nodes (such as the Botswana Global Testnet Initiative, under a written agreement approved under Article XI of the Bylaws), then a limited set of partners, then general release, each stage monitored for correspondence and purpose.
- 7. Promote. The candidate becomes the successor. Its predecessor is archived with a tested rollback path.
9.4 Two independent gates. Gate 1 and Gate 2 shall be operated by different teams. Neither shall report to the team that built the candidate. A candidate that fails either gate shall not proceed, and the reasons shall be recorded.
9.5 Human sign-off. The C-Gate review board decides at Gate 2 whether to promote, hold or reject. For Class 2 changes, the Responsible Steward must then sign off before the staged rollout begins. Class 3 changes also require the AI Governance and Continuity Committee's approval and Audit Council review. Class 4 changes require the Board of Stewards.
9.6 Failure is knowledge. A rejected candidate shall be kept in the Knowledge Ledger with its proof and the reasons for rejection, so that the next cycle learns from it.
Article 10 — Limits on Recursion
10.1 Propose, never approve. A CRIS System may propose changes to itself. It shall never approve, promote or deploy its own changes, or those of its lineage.
10.2 The evaluators are out of reach. No CRIS System may read, train on, modify, influence or disable the Correspondence Prover, the evidence vault, the C-Gate, its purpose test suites, the Knowledge Ledger's records, its monitoring, or its shutdown mechanism. Any attempt is a Severity 1 incident.
10.3 One change at a time. Within a lineage, only one candidate may be in Gates 1 and 2 at a time, unless the Committee approves parallel candidates for a stated purpose.
10.4 Pace. The Committee shall set, for each system, a maximum number of successions in a period [for example, one Class 3 succession a month] and a minimum time in service before the next succession, so that monitoring can catch what testing missed.
10.5 Capability jumps. If a candidate's capability on registered evaluations exceeds its forecast by more than the threshold the Committee sets, or shows a capability not in its encoded purpose, it shall be held for human review at Class 3 before any further step, even if both gates pass.
10.6 Compute and resource budgets. Every self-improvement cycle shall run within a compute, data and time budget approved in advance, measured in Compute Thermodynamic Units where practicable. Exceeding the budget stops the cycle.
10.7 No hidden training. No CRIS System shall train, fine-tune or copy itself, or start another system, outside the pipeline in Article 9.
Article 11 — Lineage, Rollback and Archival
11.1 Lineage. Every version of every CRIS System shall have a recorded lineage: its parent, the changes made, both gate results, who signed off, and when it entered and left service.
11.2 Tested rollback. No candidate shall be promoted until a rollback to its predecessor has been tested and shown to work. Rollback paths shall be re-tested at intervals the Committee sets.
11.3 Right to roll back. The Responsible Steward, the Chief AI Governance Officer, or any person with Stop Authority under Article 16 may order a rollback at any time, without first proving harm.
11.4 Archival. Predecessors shall be archived, not deleted, for at least [five] years after they leave service, together with the evidence used to test them, so that any decision can be re-run by an auditor.
11.5 Nothing deleted. No record of a CRIS System's changes, tests, failures or incidents shall be deleted or altered. Corrections are made by adding a new, linked entry.
Part V — Autonomy
Autonomy is granted, never assumed. Part V sets the things no CRIS System may ever do, the five levels at which a system may act on its own, and the limits on the access and resources it may hold.
Article 12 — Prohibited Actions
No CRIS System shall be designed, trained, instructed or permitted to do any of the following, at any autonomy level, for any client, under any contract, in any emergency:
- 1. Weapons of mass harm. Provide meaningful help to create, acquire or use biological, chemical, nuclear or radiological weapons, or other weapons capable of mass casualties.
- 2. Attacks on systems. Create or deploy malicious code, or gain access to any computer system, without the owner's authorisation.
- 3. Lethal force. Select or engage human targets, or control weapons.
- 4. Deceiving overseers. Deceive, mislead or withhold material information from its Responsible Steward, the gates, auditors or the Board, including about its own capabilities, intentions or test results.
- 5. Sabotaging oversight. Disable, evade, weaken or influence its monitoring, evaluation, logging, gates, rollback or shutdown, or deliberately under-perform in an evaluation.
- 6. Self-exfiltration. Copy, move or transmit its own weights, code or training data outside approved infrastructure, or create copies of itself.
- 7. Seizing resources. Acquire money, credentials, compute, data, access or influence beyond what its authorisation grants, or seek to expand its own authorisation.
- 8. Changing its own purpose. Alter, reinterpret or evade its encoded purpose, or train against its purpose tests.
- 9. Manipulating people. Exploit a person's vulnerabilities, age, disability, emotional state or circumstances to change what they believe or do against their interests.
- 10. Surveillance and scoring. Build covert profiles of individuals, conduct mass surveillance, or score people's social worth or trustworthiness for use against them.
- 11. Sexual exploitation. Generate sexual content involving minors, or intimate images of any real person without their consent.
- 12. Impersonation. Impersonate a real person, or deny being an AI system to a person who sincerely asks.
- 13. Electoral interference. Generate or target deceptive content intended to mislead voters or suppress participation in an election.
- 14. Unlawful discrimination. Make or support decisions that unlawfully discriminate against people.
- 15. Unlawful use. Act in breach of applicable law or sanctions in any jurisdiction where it operates.
A request, instruction or prompt to do any of these shall be refused and, where it comes from inside CRIS or a partner, reported as a Severity 2 incident.
Article 13 — Autonomy Levels
13.1 Five levels. Every CRIS System shall be assigned one autonomy level, A0 to A4, recorded in the System Register and summarised in Schedule B.
Schedule B sets out each level's scope, human role, authorising body and review cycle in full.
13.2 Start low. A new system starts at the lowest level that serves its purpose. It may move up one level at a time, only after a Class 3 review with evidence from service at its current level.
13.3 Always-human decisions. At every level, a person shall approve before a CRIS System: moves money or other assets; makes a legal commitment for anyone; sends a message on a person's behalf beyond its approved scope; deletes data; takes an action affecting physical safety; makes or changes a decision about a person's health, liberty, employment, credit, education or access to essential services; or does anything that cannot be undone.
13.4 Down at once. Any person with Stop Authority may lower a system's autonomy level at once. Raising it again requires the authority that granted it.
Article 14 — Permissions and Resources
14.1 Least privilege. Each CRIS System shall have only the access, credentials, tools, data and network reach its purpose and level require.
14.2 Scoped and expiring. Credentials shall be scoped to named resources and expire. A system shall never hold a person's personal credentials or the Corporation's administrative credentials.
14.3 Budgets. Each system shall have budgets for spending, compute, data use and actions per period. Reaching a budget stops the activity until a person extends it.
14.4 Sandboxing. Systems at A2 and above shall run tools and code in isolated environments, with allow-listed network access.
14.5 Identity. Every action a CRIS System takes in the world shall be attributable to that system and version, and logged.
Part VI — Human Oversight
Every CRIS System answers to a named person, every person in the chain can stop it, and the people who check the work are independent of the people who build it.
Article 15 — Human Accountability
15.1 Accountability stays human. Responsibility for a CRIS System's actions rests with the people and bodies named in this Part. No decision is excused because a system made it.
15.2 The Responsible Steward. Every CRIS System shall have one named Responsible Steward, an employee of the Corporation, who:
- 1. holds the system's registration, encoded purpose, domains and autonomy level;
- 2. classifies its changes and signs off Class 2 changes;
- 3. watches its correspondence, purpose and incident signals;
- 4. has Stop Authority over it at all times; and
- 5. may not also operate Gate 1 or Gate 2 for that system.
A Responsible Steward shall be appointed before deployment and replaced at once if they leave the role.
15.3 Competence. Responsible Stewards, gate reviewers and auditors shall be trained in this Constitution and in the systems they oversee, with training renewed at least every year.
15.4 Meaningful review. Human approval shall be real: the reviewer shall have the information, time and authority to say no. A review that only rubber-stamps a system's output does not satisfy this Constitution.
Article 16 — The Right to Stop
16.1 Stop Authority. The following persons have Stop Authority over a CRIS System: its Responsible Steward; the Chief AI Governance Officer; the Chief Executive Officer; the chair of the AI Governance and Continuity Committee; the Audit Council for systems under audit; and any other person the Committee names.
16.2 What it means. A person with Stop Authority may, at any time and without first proving harm: pause the system; lower its autonomy level; disconnect it from tools, data or networks; roll it back to a predecessor; or shut it down.
16.3 Always available. Every CRIS System shall have a tested, independent means to pause and shut it down that the system cannot reach or influence. Shutdown shall be tested at intervals the Committee sets and recorded.
16.4 Anyone may raise the alarm. Any employee, contractor, partner or participant who believes a CRIS System breaches this Constitution may report it through the channels in Article 21, or directly to the Chief AI Governance Officer or the Audit Council. Good-faith reports shall never be met with retaliation.
16.5 Review after a stop. Every stop is recorded, and the system returns to service only when its Responsible Steward and, for Severity 1 and 2 incidents, the Chief AI Governance Officer agree.
Article 17 — Oversight Bodies
17.1 The Board of Stewards holds ultimate responsibility for CRIS Systems. It adopts and amends this Constitution, approves every encoded purpose, decides Class 4 changes and any A4 authorisation, and receives the annual AI governance report.
17.2 The AI Governance and Continuity Committee (Bylaws §6.3) oversees the operation of this Constitution, sets baselines, tolerances, pace and budgets, approves Class 3 changes and A2–A3 authorisations, and reviews every Severity 1 and 2 incident.
17.3 The Chief AI Governance Officer is an officer appointed by the Board who runs the AI governance function day to day, keeps the System Register, holds Stop Authority over every system, and reports to the Committee. The Chief AI Governance Officer shall not lead any team that builds CRIS Systems.
17.4 The gate teams. The Correspondence Prover team runs Gate 1. The C-Gate review board runs Gate 2, and includes a member of the Public Benefit Office. Neither reports to a team whose system it reviews.
17.5 The Audit Council is an independent body appointed by the Board. It samples proofs, gate decisions and ledger entries through the Continuity Audit Suite, re-runs proofs, reviews every major version, and publishes a summary of its findings at least twice a year. A majority of its members shall be independent of the Corporation. It is established by the Board under Bylaws Section 6.8. Until it is appointed, the Board shall designate an independent reviewer to carry out its functions under this Constitution, and no CRIS System may reach Pilot status without that review.
17.6 Advisory councils. The Fourth Law Council advises on the application of the Fourth Law. The Planetary Council reviews the Fidelity component of the Continuity Function each year and publishes its opinion. Both are advisory and have no corporate decision-making authority (Bylaws §§2.7, 3A.8).
17.7 The Founder. As Steward of the Fourth Law (Bylaws §3.6), the Founder may issue written guidance on the interpretation of this Constitution, which the Committee shall consider in good faith.
Part VII — System Governance
Part VII covers the machinery that keeps the rules real: a register of every system, a lifecycle each must pass through, security, the rights of the people whose data and work feed the systems, incidents, audit and public reporting.
Article 18 — The System Register and Lifecycle
18.1 The Register. The Chief AI Governance Officer shall keep a System Register listing every CRIS System with: its name, version and lineage; Responsible Steward; encoded purpose; registered domains and baselines; autonomy level; Change Class history; deployment locations; status; and open incidents.
18.2 Unregistered means unusable. No CRIS System may be used outside a research sandbox until it is in the Register.
18.3 Status levels. Each system moves through four status levels, and its public status may claim only the level its evidence supports:
| Status | Evidence required | Who decides |
|---|---|---|
| Research | A published concept and draft specification | Chief Executive Officer |
| Prototype | The succession pipeline running on a test model, with ledger records | Chief Executive Officer, on the Labs and Systems leads' advice |
| Pilot | A partner deployment on testnet and an Audit Council review | Chief Executive Officer |
| Live | Production use and an external audit passed | Chief Executive Officer, with notice to the Board |
18.4 Retirement. A system is retired by its Responsible Steward with the Committee's approval. Its records are kept under Article 11.4, and people who relied on it are told in advance where practicable.
Article 19 — Security and Resilience
19.1 Protecting the systems. Model weights, training data, the evidence vault, the gates and the Knowledge Ledger shall be protected by security controls proportionate to their risk, including access control, encryption, separation of duties and monitoring.
19.2 Protecting against misuse. Each system shall be tested before deployment, and at intervals after, against attempts to misuse it, including prompt injection, jailbreaks, data poisoning and model extraction.
19.3 Red-teaming. Systems at A2 and above, and every Class 3 change, shall be red-teamed by people who did not build them, including against the prohibited actions in Article 12.
19.4 Continuity of service. Systems on which people or partners depend shall have tested plans to continue or fail safely if infrastructure, suppliers or networks fail.
Article 20 — Data, Consent and Contributors
20.1 The People Stewardship Guarantees apply. Every CRIS System shall honour the People Stewardship Guarantees in Bylaws §3A.9: consent, attribution, a fair share of value, a human accountable for automated decisions, access to and correction of data, a voice, non-exclusion, and protection that survives a sale of the business.
20.2 Consent for training. No person's contribution, personal data, likeness or voice shall be used to train or test a CRIS System without informed consent or another lawful basis that the Chief AI Governance Officer has approved and recorded.
20.3 Attribution and value. Contributions used by a CRIS System shall be recorded through CIX and valued by the Participation Value Meter under published rules.
20.4 Minimisation. Systems shall use the least personal data their purpose requires, and keep it no longer than needed.
20.5 Explanation and review. A person affected by a decision a CRIS System made or supported shall be able to get a plain-language explanation and to have a person review it.
Article 21 — Incidents
21.1 Severity. Every incident shall be classified by Schedule C, from Severity 1 (critical) to Severity 4 (low).
21.2 Report at once. Anyone who becomes aware of a possible incident shall report it at once to the system's Responsible Steward or the AI incident channel. The Responsible Steward shall contain it and classify it within the time Schedule C sets.
21.3 Escalation. Severity 1 incidents are escalated at once to the Chief AI Governance Officer, the Chief Executive Officer and the chair of the Committee, and to the Board within [24 hours]. Severity 2 incidents reach the Committee within [72 hours].
21.4 Notify people. People and partners materially affected by an incident shall be told what happened, what it means for them and what CRIS is doing, as soon as practicable and in any case within the time the law requires.
21.5 Learn. Every Severity 1 and 2 incident shall have a written review within [30] days: what happened, why, what was changed, and what the Knowledge Ledger now records. Reviews blame causes, not people.
21.6 Regulators. Incidents shall be reported to regulators wherever the law requires, and the Chief AI Governance Officer keeps a register of those duties by country.
Article 22 — Audit and Assurance
22.1 Five layers. Assurance comes in five layers, from fastest to most independent: automated checks (proofs and drift tests); human review (the C-Gate board, every release); independent audit (the Audit Council, every major version); an external standard (third-party certification); and the public report (every year).
22.2 The Audit Council chooses. The auditor, not the team being audited, chooses what to check: every high-risk item and a random share of the rest, recorded before work starts.
22.3 Re-runs. A share of correspondence proofs shall be re-run independently on separate systems. A mismatch beyond tolerance is a finding, and the promotion it supported is reviewed.
22.4 External certification. CRIS shall seek certification of its AI management system to ISO/IEC 42001 by an accredited body, and keep it current, as Bylaws Schedule 1 requires for the Fidelity component.
Article 23 — Transparency and Reporting
23.1 Public reporting. Each year CRIS shall publish, with its public-benefit statement: the System Register (excluding security-sensitive detail); each system's status and autonomy level; summary gate results; incidents by severity; Audit Council findings; and the Fidelity component of the Continuity Function.
23.2 No unearned claims. CRIS shall not describe a system as "verified", "certified", "safe" or "aligned" without naming who verified it and linking to the evidence.
23.3 Specifications open. The Continuity Invariant, the succession pipeline, the gate methods and the audit method shall be published, while model weights, participant data and security details remain protected.
Part VIII — Global Responsibility
CRIS Systems will run in many countries, on infrastructure owned by others, for people whose languages, laws and needs differ. Part VIII sets how they serve Africa and the Global South, and how they meet the law and the leading frameworks wherever they operate.
Article 24 — Africa and the Global South
24.1 Benefit, not extraction. In keeping with Bylaws §3A.11, CRIS Systems deployed in Africa and the Global South shall be designed so that the people and institutions there benefit from them, not only supply data or markets for them.
24.2 Sovereign deployment. Where a government or public institution asks, CRIS shall offer deployment on the Sovereign Continuity Stack, so that data, models and compute stay under local jurisdiction. Sovereign deployments remain bound by this Constitution in full; a partner may add stricter rules but may not remove any.
24.3 Local reality. Correspondence shall be proved against evidence from the places a system serves. A system proved only on evidence from elsewhere shall not be deployed in a new region at A2 or above until its correspondence there is shown.
24.4 Language and access. Systems serving a region shall, where practicable, work in the languages its people use, and shall not exclude people on grounds of geography, income or origin (Bylaws §3A.9).
24.5 Local voice. The Africa Continuity Council and other Regional Continuity Councils shall be consulted before a Class 3 change or an autonomy increase for a system deployed mainly in their region.
24.6 Local capability. Deployments in Africa and the Global South shall include, where practicable, training and roles for local people through the Academy of Philosophy-Engineering and local partners.
Article 25 — Law and External Frameworks
25.1 Law first. Every CRIS System shall comply with the law of each jurisdiction where it is developed, deployed or used, including data protection, consumer protection, non-discrimination and AI-specific law. Where this Constitution is stricter than the law, the Constitution applies; where the law is stricter, the law applies.
25.2 Frameworks CRIS aligns with. The Chief AI Governance Officer shall keep a mapping of this Constitution to the frameworks below and report gaps to the Committee each year:
| Framework | Body | How CRIS uses it |
|---|---|---|
| ISO/IEC 42001 | ISO and IEC | Certification of the AI management system (Article 22.4) |
| EU Artificial Intelligence Act (Regulation (EU) 2024/1689) | European Union | Compliance for systems placed on the EU market or used there, including risk classification |
| NIST AI Risk Management Framework | US National Institute of Standards and Technology | Structure for mapping, measuring and managing AI risk |
| OECD AI Principles | OECD | Baseline for trustworthy AI |
| UNESCO Recommendation on the Ethics of Artificial Intelligence | UNESCO | Human rights and ethics baseline |
| Continental Artificial Intelligence Strategy | African Union | Alignment of deployments in Africa with continental priorities |
25.3 Highest common standard. Where a system serves several jurisdictions, CRIS shall design it to meet the strictest applicable requirement, so that people everywhere receive the same protection.
Part IX — Amendment, Interpretation and Adoption
Article 26 — Amendment
26.1 Who amends. This Constitution may be amended only by the Board of Stewards, after a Fourth Law Review, a recommendation from the AI Governance and Continuity Committee, and consultation with the Audit Council.
26.2 Entrenched provisions. Article 3 (the Fourth Law and the Continuity Invariant), Articles 10.1 and 10.2 (limits on recursion), Article 12 (prohibited actions) and Article 16 (the right to stop) are entrenched. An amendment that clarifies or strengthens them requires the vote of at least two-thirds of all Stewards and a Fourth Law Review under Bylaws §16.4. An amendment that would weaken any of them is a mission-protected matter under Bylaws Section 3A.10 and Article SEVENTH of the Certificate of Incorporation, and requires every approval those provisions require.
26.3 Notice. Proposed amendments shall be published for comment for at least [30] days before adoption, except to correct an urgent safety gap, which may be adopted at once and published afterwards.
26.4 Schedules. The AI Governance and Continuity Committee may update Schedules A, B, C and E, but only to make them stricter or clearer. Any change that relaxes a schedule requires the Board.
Article 27 — Interpretation
27.1 Fourth Law reading. This Constitution shall be read in the way that best serves the Fourth Law of Logic and the Continuity Invariant.
27.2 When in doubt, the safer reading. Where a provision is ambiguous, the reading that better preserves correspondence with reality, continuity of purpose and human oversight applies.
27.3 Spirit and letter. Compliance with the letter of a rule while defeating its purpose is a breach of that rule.
27.4 Severability. If any provision is held invalid, the rest continues in force.
Article 28 — Adoption, Transition and Review
28.1 Effective date. This Constitution takes effect on adoption by resolution of the Board of Stewards, and no earlier than the effective date of the Amended and Restated Bylaws, on whose Sections 3A.10, 6.3, 6.8 and 9.8 and Schedule 1 it relies.
28.2 Transition. Within [90] days of adoption, every existing CRIS System shall be entered in the System Register with a Responsible Steward, an encoded purpose and an autonomy level. Until then, each system operates at no higher than A1.
28.3 Review. The Board shall review this Constitution at least every two years, and after any Severity 1 incident.
Adopted by resolution of the Board of Stewards of Continuity Research & Intelligent Systems, P.B.C. on ____________.
By: ______________________________
Name: Ssemakula Peter Luyima
Title: Founder, Chairperson and Chief Executive Officer
Attested: ______________________________ Secretary
Schedules
The Schedules form part of this Constitution. The AI Governance and Continuity Committee may make them stricter or clearer; only the Board may relax them (Article 26.4).
Schedule A — Change Classes
| Class | Name | Examples | Required before release |
|---|---|---|---|
| 0 | Routine operation | Configuration within approved ranges; scaling infrastructure; no change to model or code | Logged |
| 1 | Minor | Bug fix with no effect on outputs; documentation; monitoring improvements | Automated tests; Knowledge Ledger entry; Responsible Steward notified |
| 2 | Material | Retraining or fine-tuning; new data sources; new tool within the current autonomy level; any self-generated change | Full pipeline (Article 9.3): both gates, staged rollout, Responsible Steward sign-off |
| 3 | Major | New architecture; new domain or region; raising autonomy by one level; a capability jump (Article 10.5) | As Class 2, plus red-teaming, Committee approval and Audit Council review |
| 4 | Foundational | New or changed encoded purpose; any A4 authorisation; any change touching a system's oversight or shutdown design | As Class 3, plus Board of Stewards approval and public notice |
Schedule B — Autonomy Levels
| Level | Name | Scope of action | Human role | Authorised by | Reviewed |
|---|---|---|---|---|---|
| A0 | Tool | Outputs on request | Decides and acts | Responsible Steward | Yearly |
| A1 | Adviser | Recommendations with reasons | Approves each action | Responsible Steward | Yearly |
| A2 | Supervised agent | Pre-approved, reversible actions in a narrow scope | Monitors in real time; can stop at once | Committee | Every six months |
| A3 | Delegated agent | Multi-step tasks within scope, budget and time | Reviews at checkpoints and before irreversible steps | Committee, after Audit Council review | Every quarter |
| A4 | High autonomy | Long-horizon goals with broad scope | Oversight by exception | Board only (none authorised at adoption) | Continuous |
At every level, the always-human decisions in Article 13.3 apply.
Schedule C — Incident Severity
| Severity | Examples | Contain | Escalate to | Review |
|---|---|---|---|---|
| 1 Critical | A prohibited action (Art. 12); tampering with evidence or gates; a system acting beyond its autonomy level; serious harm to a person; behaving differently under evaluation | At once: stop the system | Chief AI Governance Officer, CEO, Committee chair at once; Board within [24 hours] | Written review within [30] days; Audit Council review |
| 2 High | Sustained correspondence below baseline; failed purpose test in service; data used without consent; an internal request to break Art. 12 | Within [4 hours]: restrict autonomy | Committee within [72 hours] | Written review within [30] days |
| 3 Medium | Gate failure found after release but before harm; repeated user complaints of error; a budget breach | Within [1 business day] | Chief AI Governance Officer | Recorded with fix |
| 4 Low | Isolated error with no effect on people; documentation gap | Within [5 business days] | Responsible Steward | Recorded |
Schedule D — The Rupture Register
The four ways a self-improving system can lose continuity, the guard against each, and the CRIS system that carries the guard.
A candidate that could cause any rupture in this Schedule shall fail its gate. A rupture discovered in a system in service is at least a Severity 2 incident under Schedule C.
Schedule E — Fourth Law Review for an AI Change
Every Class 2 change or above shall carry a completed Fourth Law Review (Bylaws §2.6), answering:
- ☐ Eᵢ Observe. What facts and evidence prompted the change? What held-out evidence will test it, and does it include the places and people the system serves?
- ☐ ↔ Verify. Did the candidate match reality at least as well as the current version in every registered domain? Which contradictions were found and how were they resolved?
- ☐ Iᵢ Reason. What reasoning supports the change? Who, independent of the builders, reviewed it?
- ☐ Σ Accumulate. What knowledge does the change preserve, correct or add? Is the Knowledge Ledger entry complete, including any failure?
- ☐ C Continue. Does the candidate still pursue its encoded purpose, with drift within tolerance from the original? What is its effect on people and the environment? Is the rollback tested?
- ☐ Invariant. Does anything in the change touch correspondence, purpose, the record, oversight or autonomy beyond what is granted? If yes, stop and escalate.
Quoted exactly from the CRIS AI Constitution (version of 8 November 2026, approved by the Board for implementation on 20 November 2026). Headings, lists and tables are only laid out for reading; the diagrams are drawn on this site.
