Sovereign Continuity Stack
Continuity-bound AI on your own infrastructure, with your own data.
In one paragraph
The Sovereign Continuity Stack lets a country or institution run CRIS's continuity-bound AI on its own infrastructure, with its own data. It packages the whole Continuity Recursion Loop, from observation to the C-Gate, to run on a partner's own hardware or a hosting provider such as LOGOS-C, under a written agreement. It tests every new version first on testnet nodes, such as those of the Botswana Global Testnet Initiative, under a written agreement, and measures the energy and cost of every improvement in Compute Thermodynamic Units (CTU). Local operators keep control, with the tools to audit the system themselves.
It adds no new operator to the loop: it is the ground the loop stands on. The other nine systems describe how the loop learns without rupture; the Stack decides where it runs and who controls it. It is how Africa and the Global South can take part in AI as architects, not only as users of systems built elsewhere. The Stack is described here as designed: no sovereign deployment has yet taken place.
Where it sits
The Stack does not carry one stage of the loop. It sits beneath all six, and beneath every other CRIS system.
every stage
Observe
Observe the world as it is, and record where each observation came from.
Verify
Verify every observation, and every change a system makes to itself, against evidence it has not seen.
Reason
Reason only from what has been verified, in steps someone else can trace.
Accumulate
Accumulate what was verified, so knowledge adds up instead of being overwritten.
Benefit
Benefit people and the environment: participants share in the value they create, and every change is weighed against our public benefit purpose.
Continue
Continue, changing anything except correspondence with reality and continuity of purpose.
Select a tile to see what it does. Every tile links to its page.
Rules for this system are proposed in the AI Constitution, Article 24.2: sovereign deployment. They are approved by the Board for implementation.
What the Stack contains
The Stack is the CRL Engine, packaged with everything a partner needs to run it locally.
- Local observation sourcesNational data, institutions and CIX participants in the country
- Local controlPartner operators run, pause and roll back; keys held locally
- Audit toolsThe Continuity Audit Suite, for local operators and regulators
- Sovereign computeThe partner’s own hardware, or a hosting provider such as LOGOS-C, under a written agreement
- Testnet nodesWhere first cycles and new versions run before release (for example BGTI, under a written agreement)
- CTU meteringThe energy and cost of every cycle, in Compute Thermodynamic Units
| Layer | What it is |
|---|---|
| The loop | The CRL Engine with its five operator systems; CIX and the Value Meter where the partner chooses |
| Local observation sources | National data, institutions and CIX participants in the country |
| Local control | Partner operators run, pause and roll back the system |
| Audit tools | The Continuity Audit Suite, for local operators and regulators |
| Sovereign compute | The partner's own hardware, or providers such as LOGOS-C, under a written agreement |
| Testnet nodes | Testnets, such as the Botswana Global Testnet Initiative's under a written agreement, where new versions run in staged deployment before release |
| CTU metering | A measure linking intelligence to its energy and compute cost |
LOGOS-C and the Botswana Global Testnet Initiative are each an organisation also led by our Founder. Any work with CRIS is under a separate written agreement on arm's-length terms; none is required to run the Stack. See related organisations.
- LOGOS-CCould host the loop, as one hosting option: sovereign, AI-native compute
- Botswana Global Testnet InitiativeCould test it: testnet nodes where new versions run first
- Academy of Philosophy-EngineeringCould train the people who run and audit it
How a deployment works
Every deployment starts with a request from a partner nation or institution and follows five steps.
- 1AssessData residency, law and infrastructure readiness
- 2DeployThe CRL Engine on partner hardware or a hosting provider such as LOGOS-C, under a written agreement, with local observation sources
- 3First cyclesOn testnet nodes, such as the Botswana Global Testnet Initiative’s, under a written agreement
- 4Hand overAudit tools to local operators; training through the Academy
- 5ReportCTU energy and cost beside capability results
- Assess data residency, law and infrastructure readiness.
- Deploy the CRL Engine on local hosting, such as partner hardware or LOGOS-C under a written agreement, with local observation sources.
- Run the first cycles on testnet nodes (such as the Botswana Global Testnet Initiative, under a written agreement).
- Hand over the audit tools to local operators, and train them through the Academy of Philosophy-Engineering.
- Report CTU energy and cost figures alongside capability results.
Assessment and staging
Before anything is installed, three questions are answered: what data must stay in the country, what the law requires, and whether the infrastructure and people are ready. The decision to deploy in a new country is the CEO's, with Sovereign Deployment, legal counsel and local partners consulted, and this assessment as the evidence.
- Which data must stay in the country
- Where the ledger and backups live
- What may cross borders, and on whose consent
- Data protection law and the regulator
- AI, procurement and export rules
- Who may audit, and how findings are published
- Power, cooling and connectivity
- Hardware: LOGOS-C or the partner’s own
- Local operators and their skills
A new country then follows the same staging as every CRIS release: first in a sandbox, then on testnet nodes (such as the Botswana Global Testnet Initiative, under a written agreement), then with a limited set of partners. Each move needs the previous stage's correspondence proofs, a C-Gate decision and a tested rollback. A failure sends the deployment back one stage.
- SandboxCRIS lab; synthetic and public data
- Testnet nodesFor example BGTI, under a written agreement; first cycles, proofs and C-Gate decisions run end to end
- Limited partner useLocal observation sources; named users
- Partner operationLocal operators; rollback path tested
Sovereignty and data residency
Two of our design principles decide this. No CRIS system assumes a single foreign cloud: each can run on partner infrastructure. And we publish our interfaces, the Continuity Invariant and the audit method, while model weights and participant data are protected. Our proposal is that participant data, local observations and the partner's own record of what the system has learned stay in the country, and that what we share openly is the method.
- Participant data and consent records
- Local observations and evidence
- The partner’s Knowledge Ledger
- Operational keys and logs
- Interfaces and specifications
- The Continuity Invariant
- The audit method
- The CTU reporting format
- Model updates between deployments
- Aggregate results for the benefit report
- Incident details across partners
Measured cost: CTU
Every CRIS system reports its compute and energy in Compute Thermodynamic Units, so that gains in capability can be weighed against their cost. For a sovereign deployment, this means each partner sees the compute and energy its own deployment used. Our benefit measures carry it into the public benefit report through one environment indicator, whose exact form is still being chosen, with deployments running on local infrastructure with local data reported beside it. See CTU on this site. No figures are shown until they are measured.
- Each cycleProposal, proof, gate decision
- MeteredCompute, energy and cost in CTU
- ComparedCorrespondence gain per CTU
- ReportedTo the partner, and in the benefit report
Scorecard: Environment
One environment indicator, in CTU; its exact form is being chosenScorecard: Sovereignty
Deployments running on local infrastructure with local dataLocal control and audit
Local operators run the system and can pause it or roll it back. They receive the same audit tools our independent reviewers use, and training through the Academy of Philosophy-Engineering. No partner receives a stack that has not passed an external audit.
- Local operatorsRun cycles, pause and roll back; hold operational keysCannot promote a version that has not passed the C-Gate
- Academy of Philosophy-EngineeringWould train local staff, under a written agreement (none recorded yet)Cannot sign off its own audit
- Continuity Audit SuiteGives operators and regulators the tools to check proofs, decisions and the ledgerIs not run only by CRIS staff
- Audit Council (once appointed) and external auditorCheck the Stack before handover and afterNot managed by the Sovereign Deployment team
- CRIS Sovereign DeploymentSupports and updates under the partner agreementCannot override local control outside the agreement
- Partner regulatorOversees under local law—
| Party | Does | Cannot |
|---|---|---|
| Local operators | Run cycles, pause and roll back; hold operational keys | Promote a version that has not passed the C-Gate |
| Academy of Philosophy-Engineering | Would train local staff, under a written agreement (none recorded yet) | Sign off its own audit |
| Continuity Audit Suite | Gives operators and regulators the tools to check proofs, decisions and the ledger | Be run only by CRIS staff |
| Audit Council (once appointed) and external auditor | Check the Stack before handover and after | Be managed by the Sovereign Deployment team |
| CRIS Sovereign Deployment | Supports and updates under the partner agreement | Override local control outside the agreement |
| Partner regulator | Oversees under local law | — |
For partner nations
Our plan aims for the first sovereign deployment, passing an independent audit, in 2029. You can prepare before then: register interest, join testnet cycles (such as the Botswana Global Testnet Initiative's, under a written agreement) from 2027, train your staff through the Academy and take part in the CIX pilot. These dates are proposals, not commitments.
- 2027 · Prove the loopInvariant v1; partners can join BGTI testnet cycles; Academy training; CIX pilot
- 2028 · Open the exchangeFirst benefit report audited against the standard
- 2029 · Go sovereignFirst sovereign deployment passes an independent audit
- 2030 · CompoundThe board reviews the next five-year plan
- Register interestPartner with us, or read Sovereign AI
- BriefingScope and readiness questions
- Testnet cyclesJoin testnet cycles (for example BGTI’s, under a written agreement), from 2027 (proposed)
- Train staffThrough the Academy of Philosophy-Engineering
- Assessment and agreementThen the five-step workflow
Status and read more
This status is set by the CRIS team and links to the evidence behind it. Until a full test cycle runs on sovereign compute with CTU metering, the Stack remains a design.
Current status: Status to be confirmed
| Level | Evidence required | Who sets it |
|---|---|---|
| Research | Concept page; deployment workflow and readiness checklist published | CEO |
| Prototype | The CRL Engine running a full test cycle on LOGOS-C in the lab, with CTU metering | CEO, on the Labs and Systems leads’ advice |
| Pilot | First partner testnet cycles on Botswana Global Testnet Initiative nodes, with a signed partner agreement | CEO |
| Live | A sovereign deployment that has passed an independent audit and runs under local control | CEO, with notice to the board |
For developers: the Deploy path, the staged rollout and the CTU metering method, drafts 0.1.
- Sovereign AI, the Systems overview, the Continuity Audit Suite and our ecosystem.
- Recursion Without Rupture: Chapter 11.4 (LOGOS-C and the reference architecture), Chapter 15 (sovereign compute in Africa and the Global South) and Chapter 18.4.
- Related systems: CRL Engine, C-Gate, Knowledge Ledger, Continuity Audit Suite, CIX.
