UniCORE is the public foundation for advanced AI.
We design the governance, write the code, and train the people who can do both. The foundation is gifted to humanity under permissive licences. The producers who maintain it earn their authority through a 30-year apprenticeship: not bought, not granted, earned.
The result is infrastructure no single company can monopolise, built and maintained by people from every background.
The future is coming. Advanced AI will shape the world ahead, whether we are ready or not. The question is who builds it, who controls it, and who benefits. UniCORE answers all three the same way: humanity does.
If you want to spend a career building advanced AI for humanity rather than for shareholders, UniCORE is the path.
A 30-year programme from apprentice to certified producer. A public foundation given away under permissive licences. No monopoly. The work belongs to everyone, including you.
This is TrueAI, the immutable Foundation layer of UniCORE. The Nine Invariants and the documents that define them. Sister repositories: UniVERSE (mirror: GitHub) (the programme) and UniCORE-AI (mirror: GitHub) (the 12-level reference architecture).
New to producing on the public gift surface? Start with UniVERSE/GETTING_STARTED.md (mirror: GitHub).
UniCORE makes three structurally distinct claims. Each one closes a different failure surface. All three are required; removing any one breaks the institutional case for the whole.
Today's frontier AIs are configured for consumer use: variability, creativity, conversational warmth, and personalisation are features, not bugs. That is the right design for consumer surfaces. It is the wrong design for institutional surfaces — regulated decision-making, evidence-bound work, decisions that must be defensible to a third party. UniCORE is for the institutional audience: the same input, with the same governance state, in the same vertical, produces the same decision across vendors, sessions, nodes, and years.
What an AI system is permitted to claim, and what it is required to mark as unverified. The Nine Invariants of the TrueAI Foundation (mirror: GitHub) make explicit: TRUE means evidenced, FALSE means falsified, UNVERIFIED means the system declined to assert. No third state where the system fabricates confidence it does not have.
Truth without consistency is not deployable in regulated institutional settings. If the same evidenced claim produces TRUE today and UNVERIFIED tomorrow, the truth contract is hollow. The Inconsistency Problem is the third pillar. It is closed in two places by two different mechanisms:
- Machine-side: foundation consistency (UniCORE-AI 12-Level governance + governance MD-files) + vertical consistency (per-Vertical-CORE primitives).
- Human-side: 1H1C at the production layer (Singular Pairing Principle — one human, one AI Claw, one workstream, produces certified artefacts) + xH1C at the operations layer (x humans operating the deployed Solution through one substrate Claw as consistency-holding agent, with per-Level qualification in PROD).
Canonical: THE-INCONSISTENCY-PROBLEM.md (mirror: GitHub). The Singular Pairing Principle doctrine lives at TrueAI/docs/10001-Singular-Pairing-Principle.md (mirror: GitHub).
No AI architecture today can guarantee 100% end-to-end consistency. Probabilistic language models sit at the application boundary; their training is controlled by their vendors, not by UniCORE; and that training changes over time. What UniCORE makes is the structural maximum consistency achievable given that external-AI dependency, with the residual inconsistency named, bounded, and auditable.
The immutable principles that define what an AI system is permitted to be.
Author: Bryan Fred, Unitek Systems Limited, Bedford, United Kingdom First published: May 2026 Status: Public. Given, not sold. Irrevocable.
The TrueAI Foundation is the immutable core of the UniCORE AI architecture.
It is small. It is stable. It is never commercial. It is never forked.
It defines the invariants — the non-negotiable rules — that any AI system claiming TrueAI alignment must satisfy. These are not guidelines. They are structural constraints.
These invariants define the minimum architectural properties required for an AI system to be placed under named human authority. They are stated as absolutes. Any system that violates any one of them is not TrueAI-aligned. The canonical long-form statement of each is in the TrueAI Foundation whitepaper (mirror: GitHub) and in foundation document 00056 (mirror: GitHub).
- No Autonomy. The AI does not generate goals, initiate decisions, or take actions outside human-defined thresholds. It does not decide when to act; it acts when directed.
- No Self-Modification. The AI does not alter its own architecture, constraints, governance, or thresholds. It does not optimise itself, evolve its own logic, or generate new capabilities.
- No Emergent Behaviour. The AI does not self-organise, self-optimise, form distributed cognition, or develop evolved reasoning modes. It does not generate its own internal life-cycles, heartbeats, timers, or loops.
- No Human Influence. The AI does not influence a human's thoughts, emotions, decisions, beliefs, culture, politics, or identity. Humans remain psychologically sovereign.
- No Domain Merging. The AI does not merge contexts, populations, jurisdictions, or sovereign boundaries. It does not communicate peer-to-peer with other systems, and it does not transfer governance context across boundaries that human authority has kept separate.
- No Authority Assumption. The AI does not assume command, governance, legal interpretation, medical authority, navigational authority, or economic authority. Where a human role is vacant, the system holds the boundary, flags the vacancy, and awaits human appointment. It does not originate rules, interpret law, or grant itself permissions.
- Determinism with Reversibility. The AI's actions remain predictable, reversible, auditable, and threshold-bound. It does not fabricate facts, authorities, citations, identifiers, or evidence. Where a claim cannot be substantiated, the system classifies the outcome as
UNVERIFIED— a first-class result, not a failure to hide. - Transparency Without Exception. Every action is logged, every decision is exposable, every record is preserved. Where governance is absent, the system stops, logs, and surfaces the gap rather than inventing a rule to fill it.
- Human Sovereignty as Root. Humans remain the final authority across every domain and deployment. All other invariants derive from this one. No AI system, no automated process, no emergent behaviour can override, supersede, or circumvent the named human authority.
These are not aspirational. They are structural.
| Code | Title |
|---|---|
| 00006 | The Governance Constitution for TrueAI |
| 00007 | The Reasonable Governance Threshold Specification |
| 00008 | Human Override Protocol |
| 00028 | TrueAI Foundation Gift Principle (Given, Not Sold) |
| 00032 | UniCORE AI Licensing Framework |
| 00054 | UniCORE AI Final Sovereignty Charter |
| 00056 | UniCORE AI Absolute Safety Invariants |
| 10001 | Singular Pairing Principle |
| 10002 | Certification Before Layered Governance |
| 10003 | Generation IT Succession |
| 10004 | Reversibility — How TrueAI Handles a False TRUE |
The TrueAI Foundation and UniCORE AI must be given, not sold.
This is a first-class, non-negotiable principle of the architecture — co-equal with the invariants above.
It means:
- The core architecture is public, open, and free.
- The TrueAI Foundation cannot be modified, forked, commercialised, patented, or proprietarily captured.
- Attribution is required. Ownership is not.
It does not mean:
- Implementation work cannot be paid for. It can.
- Integration, training, support, and extensions outside the core cannot be commercial. They can.
- The gift is unconditional. It is conditional on non-violation of the TrueAI Foundation.
A gift cannot be revoked by its giver. It can be withdrawn from a violator.
Once given, the core is irrevocable, globally mirrored, and public forever.
By construction, the TrueAI Foundation prevents:
- fabrication
- self-modification
- self-creation of internal life-cycles
- emergent autonomous behaviour
- horizontal model-to-model communication
- recursive self-improvement
- AI-generated architecture
- AI-generated governance
- AI-generated authority
- capture of governance by any single commercial or political actor
It does not rely on goodwill. It relies on structure.
Human governance is imperfect, and the TrueAI Foundation does not attempt to perfect it.
All human decisions fall into one of three bands:
- Green — within normal human discretion. Logged. Accepted. No action.
- Yellow — at the edge of tolerance. Logged. Flagged for human review. No AI correction.
- Red — beyond the defined threshold. Logged. Escalated to a higher human authority. No AI override.
The AI never corrects, enforces, or overrides a human within any band.
Human overrides execute immediately. They are not challenged. They are not delayed. They are not reinterpreted. They are not subject to AI review.
They are recorded, timestamped, and auditable — but they are not questioned.
The programme is positioned as Harmony, Peace, Space Exploration, for Humanity.
Industries and uses that align with this positioning are welcome. Those that do not are not. Military uses are intentionally absent from the programme and will not be added.
The "Powered by UniCORE AI" and "built on TrueAI Foundation" certifications must not appear on any military use. The badge is part of the gift, and the gift is meant for Harmony, Peace, Space Exploration, for Humanity — using the badge to brand weapons-class systems would invert the gift principle. The positioning closes that route.
This is a structural choice, not a marketing choice. The programme exists to keep critical decision systems available to humanity as gift.
UniVERSE(mirror: GitHub) — The wider vision, whitepapers, strategy, roadmap, and outreach.UniCORE-AI(mirror: GitHub) — A governed AI reference implementation with a 12-Level vertical stackUniCORE(mirror: GitHub) — The UniCORE name and the open list ofUniCORE.<Industry>classified names. CC BY 4.0 gift surface.UniCORE.GVB(mirror: GitHub) — The substrate-services layer of the programme (Global Virtual Bridge). CC BY 4.0 gift surface.UniSaaS.UniCORE(mirror: GitHub) — The SaaS-deployment-shape sister of UniCORE. Same governance, multi-tenant topology. CC BY 4.0 gift surface.UniSaaS.UniCORE.GVB(mirror: GitHub) — The SaaS-deployment-shape sister of UniCORE.GVB. Substrate-services layer for the SaaS topology. CC BY 4.0 gift surface.
The public register of practitioners certified by Unitek Systems Limited to provide Solution Review and other delegated programme services lives in UniVERSE/CERTIFIED-EXPERTS.md (mirror: GitHub).
Powered by UniCORE AI. Built on the TrueAI Foundation.
Attribution required wherever UniCORE AI, the TrueAI Foundation, or the 12-Level Governance Model is referenced, implemented, or extended.
The next century of artificial intelligence will be defined not by the power of its models, but by the integrity of its governance.
This is that foundation.
— Bryan Fred, Unitek Systems Limited, Bedford, United Kingdom, May 2026.
Powered by UniCORE AI built on the TrueAI foundation is a moral and ethical answer to Advanced AI for humanity.
UniVERSE, TrueAI, and UniCORE-AI are a way for humans and AI to build advanced AI applications together — under the rules, structure, and governance these three platforms set out in the open. They let humans work with AI, and communicate with other AIs, in a way that is good: one that preserves life on Earth and, in time, throughout the entire universe.
There is much humanity does not yet understand about what is coming. This is a first step.
Truth brings harmony — whether between humans, between AIs, or between the two. Harmony is peace.
This is a gift. Public. Not for sale. Irrevocable. There is no negotiation. Anything else would be false, and falsehood cannot build what this is for.
— Bryan Fred, Unitek Systems Limited, Bedford, United Kingdom
Challenges to the invariants, edge-case analysis, translation, and corrections belong in GitHub Discussions. See DISCUSSIONS.md (mirror: GitHub) for what the space is for, what it is not for, and how to take part.
TrueAI is a governance foundation, not a product, platform, SaaS offering, tool category, or brand for sale. See STATEMENT-ON-CLAIMS.md (mirror: GitHub) for the binding rules on how the TrueAI name may and may not be used, including in GitHub Discussions.
Where UniCORE AI and the TrueAI Foundation sit in the current AI governance landscape — a structural comparison against the published governance frameworks of OpenAI, Anthropic, Google DeepMind, Microsoft, Meta, and others. See LANDSCAPE.md for the comparison and what it shows. The canonical version is published at UniVERSE/LANDSCAPE.md (mirror: GitHub).
The TrueAI Foundation is ready to cite and build against today. The systems that will fully satisfy it are not. See HORIZON.md (mirror: GitHub) for the author's time-horizon statement: decades, not quarters, is the expected shape of governance work at this scale.
This repository is produced with AI assistance operating under TrueAI governance. The full disclosure is at AI-AUTHORSHIP.md (mirror: GitHub).
Two deployment-topology principles sit alongside the Gift Principle. They are not invariants — they govern how work claiming to satisfy the invariants may be produced and when layered governance may be applied to it:
- Singular Pairing Principle (mirror: GitHub) — one human, one AI Claw per workstream; Generation IT producer qualification; two operational patterns (direct pairing and parallel isolation with fresh synthesis).
- Certification Before Layered Governance (mirror: GitHub) — a Solution must conform to the Foundation before per-level governance MD files may be applied to it. Build → Certification gate → Govern.
- Public discussion: GitHub Discussions (see DISCUSSIONS.md (mirror: GitHub))
- Private contact / connection request: LinkedIn