Legal and charter

Global responsibility

Systems that serve the places they run in, under the strictest law that applies.

Approved by the Board for implementation · 20 November 2026

Approved

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.

In one paragraph

Summary, in plain words — not a quote

Under Part VIII, CRIS systems deployed in Africa and the Global South would have to benefit the people there. Governments could run them under local jurisdiction on the Sovereign Continuity Stack, still bound by the whole Constitution. Correspondence would have to be proved on local evidence, in local languages where practicable, with regional councils consulted. Systems would have to follow the law everywhere they operate, and the Constitution would be mapped to leading AI frameworks.

The text

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:

FrameworkBodyHow CRIS uses it
ISO/IEC 42001ISO and IECCertification of the AI management system (Article 22.4)
EU Artificial Intelligence Act (Regulation (EU) 2024/1689)European UnionCompliance for systems placed on the EU market or used there, including risk classification
NIST AI Risk Management FrameworkUS National Institute of Standards and TechnologyStructure for mapping, measuring and managing AI risk
OECD AI PrinciplesOECDBaseline for trustworthy AI
UNESCO Recommendation on the Ethics of Artificial IntelligenceUNESCOHuman rights and ethics baseline
Continental Artificial Intelligence StrategyAfrican UnionAlignment 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.

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.

In pictures

Summary, in plain words — not a quote
  • ISO/IEC 42001CRIS would seek certification of its AI management system (Article 22.4). Not held todayISO and IEC
  • EU Artificial Intelligence ActIts requirements, including risk classification, mapped for systems placed on the EU market or used thereEuropean Union
  • NIST AI Risk Management FrameworkA structure for mapping, measuring and managing AI riskUS NIST
  • OECD AI PrinciplesA baseline for trustworthy AIOECD
  • UNESCO Recommendation on the Ethics of AIA human rights and ethics baselineUNESCO
  • Continental Artificial Intelligence StrategyLining up deployments in Africa with continental prioritiesAfrican Union

The Chief AI Governance Officer would keep a mapping of the Constitution to each framework and report gaps to the Committee each year. Where a law applies, the law comes first (Article 25.1).

Frameworks CRIS would align with, Article 25.2. Alignment, not certification: a mapping shows where the rules meet a framework, not that any system meets it. A summary of the table in the text.

In practice

For builders. A system proved only on evidence from one place would not be ready for another. To act at A2 or above in a new region, it would first have to show correspondence on evidence from there. Where a system serves several jurisdictions, it would be designed to meet the strictest requirement that applies.

For Responsible Stewards. The regional council would be consulted before a Class 3 change or an autonomy increase for a system used mainly in its region. The Africa Continuity Council is the first of these councils proposed in the bylaws. Consultation is advice, not a veto: the decision stays with the bodies in Part VI.

For partners and governments. A government or public institution could ask for a sovereign deployment. Its data, models and compute would then stay under its own jurisdiction.

The whole Constitution would still apply. A partner could add stricter rules but could not remove any. The Sovereign AI page explains why this matters.

For the public. The aim is benefit, not extraction: people should gain from these systems, not only supply data or markets for them.

Systems would work in local languages where practicable. They would not exclude people because of where they live, what they earn or where they come from. Deployments would include training and roles for local people where practicable.

Law first, then frameworks

The law always applies. Where the proposed Constitution is stricter than the law, the Constitution would apply; where the law is stricter, the law would.

The frameworks in Article 25.2 are a different matter. CRIS would map its rules to them and report gaps each year.

That mapping shows where the rules and a framework meet. It is not a certificate, and it is not a legal opinion that any system meets any law or standard. For ISO/IEC 42001, CRIS would seek certification of its AI management system; none is held today.

Where it is carried out

Awaiting counsel. Counsel is confirming how the EU AI Act and the other frameworks in Article 25.2 should be described, so that alignment is not read as a legal opinion. Until then, this page describes them only as frameworks CRIS would map its rules to. Nothing here is legal advice.

Next steps

Owner: Governance and ComplianceNext review: 2027-03-01