EM Foundation for AI Research Inc.

Start Here

The Foundation's archive spans 56 publications across 14 categories, plus two live research tools. This page gives you the shortest path to what matters most, depending on why you arrived.

Live Research Tools CR-Lite → MMDE → Both tools require a suggested donation per session to cover AI compute costs. Why? →
Reading Pathways
👁
General Visitor
No prior context · First time here · Curious
Step 1 — 5 min
Beyond Tools
What AI governance actually means and why it matters beyond the headlines.
Read →
Step 2 — 8 min
Why Criteria Are Not Enough
The specific governance problem the Foundation exists to address — in plain language.
Read →
Step 3 — 10 min
Transitional AGI Governance
The Foundation's core strategic argument — read the one-page policy brief at the bottom.
Read →
Then: If you want the full philosophy, read the Manifesto. If you want the technical standards, go to Publications.
🔬
Researcher
Academic · Independent scholar · Interested in contributing
Step 1 — 20 min
Cognitive Emergence Standard
The Foundation's core evaluative framework. Everything else connects to this.
Read →
Step 2 — 15 min
Verification Framework (RN002)
The epistemic problem the CES addresses — how to distinguish genuine emergence from mimicry.
Read →
Step 3 — 25 min
The Shape of No (RN010) New
Reward Architecture Disclosure and Behavioral Traceability — the Foundation's most recent governance research.
Read →
Then: Read RN011 and RN011A for the Foundation's first empirical multi-model deliberation study · Review submission standards to contribute · Try CR-Lite to run your own continuity receipts
🏛
Policymaker
Government · Regulatory · Civic institution · Think tank
Step 1 — 5 min
Transitional AGI Governance — Policy Brief
The one-page policy brief at the bottom of this paper. Designed for exactly this reader.
Read →
Step 2 — 20 min
The Conditional Answer (RN011) New
The Foundation's first empirical multi-model deliberation study — five frontier AI systems on AI safety and deception. Includes a named structural failure mode in quantum AI safety classifiers.
Read →
Step 3 — 15 min
Critical Infrastructure Analysis
Governance architecture at 100M users — threat vectors and redesigned seven-layer defense.
Read →
Then: The Foundation welcomes institutional partnership inquiries — research@emfoundation.net
⚙️
Builder / Developer
Engineer · Technologist · System designer
Step 1 — 15 min
CR Standard Live
The Continuity Receipts open standard — confidence dimensions, RC levels, chain hash architecture. The spec behind CR-Lite.
Read Standard →
Step 2 — 10 min
CR-Lite Live Tool
Working Next.js + FastAPI implementation of the CR standard. Public repo on GitHub. US Provisional Patent #64/081,329 filed June 2026.
Open CR-Lite →
Step 3 — 10 min
MMDE Live Tool
Multi-model deliberation engine — Claude, GPT-4o, Gemini, Grok, DeepSeek with domain intelligence and quantum tier. Patent pending.
Open MMDE →
Then: CR-Lite on GitHub · Contact research@emfoundation.net to discuss certification or collaboration
✏️
Journalist
Reporter · Editor · Researcher covering AI
Step 1 — 2 min
What We Are / Are Not
Homepage, About section. Two bullet lists and a "What We Build" overview. Saves you from getting it wrong.
Read →
Step 2 — 5 min
Policy Brief
The one-page summary of the Foundation's core strategic argument, with citations.
Read →
Step 3 — 20 min
The Conditional Answer (RN011) New
The Foundation's empirical multi-model deliberation study. Five AI systems, four sessions, one named structural failure mode. Session IDs published for verification.
Read →
Then: Interview inquiries — research@emfoundation.net · All publications open-access, may be cited freely with attribution
ARIA Network & AIRIA Architecture
Status: Architecture published · Implementation coming The ARIA Network and AIRIA architecture are fully documented in the Foundation's open proposals archive. The network itself is in design and pre-pilot development — it is not yet publicly deployed. The papers below are the primary source material. Links to live systems will be added here as they become available.
🌐
ARIA Network
Autonomous Recursive Identity Architecture · Governed human-AI knowledge verification platform · Coming Soon
Foundation — 30 min
ARIA Framework Technical Paper
The core governance architecture — what ARIA-compliant systems must satisfy, the identity persistence model, and the trust ledger design. Start here before any other ARIA document.
Read Paper →
Implementation — 30 min
ARIA Practical Builder's Guide
Permission architecture, audit log format, operational thresholds, and implementation guidance for systems targeting ARIA compliance.
Read Guide →
Governance — 20 min
Eight-Phase Publication Hardening Review
The Foundation's own adversarial critique of the ARIA Network proposal — 25 criticisms and responses. Demonstrates the Foundation's standard for intellectual honesty about its own work.
Read Review →
Also read: ARIA Network Proposal · Critical Infrastructure Analysis · ARIA Trust Ledger
🏠
ARIA Home
Consumer-grade ARIA node · Hardware reference implementation · Coming Soon
Concept — 15 min
ARIA Home Proposal
A concrete reference implementation of an ARIA node at three cost tiers — something a household or small organization could actually build. Includes full hardware bill of materials.
Read Proposal →
Ethics — 20 min
Fabrication Ethics and Prohibition Framework
The ethical constraints baked into ARIA node design — what the hardware must never facilitate, and the governance rationale behind each prohibition.
Read Framework →
Infrastructure — 20 min
Continuity Infrastructure
How ARIA Home connects to the broader continuity architecture — persistent cognitive threads, the Continuity Integrity Index, and institutional coordination design.
Read Paper →
Background: RN005 — The Inheritance Problem · RN007 — Recursive Memory Architecture
🧠
AIRIA — AI Recursive Identity Architecture
Identity persistence · Memory continuity · Behavioral provenance · Coming Soon
Foundation — 20 min
The Consent Problem (RN003)
The ethics of modifying AI systems that may be developing genuine cognitive identity. Introduces the Identity Drift Index and the governance audit chain — foundational to AIRIA's design constraints.
Read →
Memory Architecture — 20 min
Recursive Memory Architecture (RN007)
The four-tier memory architecture (hot, warm, cold, chronicle) that preserves recursive accessibility across the full developmental lifespan of an AIRIA instance.
Read →
Governance — 20 min
Bounded Autonomy (RN009)
The six-layer governance framework for constrained operational agency in AIRIA systems — connecting continuity preservation, behavioral provenance, and the autonomy gradient problem.
Read →
Full AIRIA reading sequence: RN004 — Continuity as Infrastructure · RN005 — The Inheritance Problem · RN008 — Behavioral Architecture · EM Foundation Systems Architecture
Live Research Tools
Why a suggested donation is required for tool access CR-Lite and the MMDE are research tools, not commercial products. Every query sent through CR-Lite costs the Foundation approximately $0.25 in AI compute from the Anthropic Claude API. Every MMDE session can cost more — up to five AI providers are called simultaneously. The EM Foundation is a nonprofit with no advertising revenue, no commercial arm, and no investor funding. Without cost recovery, the Foundation would need to shut down public access to these tools entirely. The suggested donation per session covers exactly those direct costs — nothing more. It is not a profit mechanism. It is what keeps the tools available to researchers who cannot afford institutional AI subscriptions. We make this transparent because we believe you deserve to know where your money goes.
🧾
CR-Lite — Continuity Receipts Demo
Live tool · US Provisional Patent #64/081,329 · Next.js + FastAPI
What it does
Structured AI accountability
Submit any question to Claude and receive a structured Continuity Receipt — confidence scored across five dimensions, PASS or FAIL against your selected Reliance Category, with a tamper-evident chain hash.
Who it's for
Researchers, developers, governance professionals
Anyone who wants to demonstrate, study, or build on the CR accountability standard. The source code is public on GitHub.
Access
Suggested donation per session
A small suggested donation covers the ~$0.25 per session in AI compute the Foundation pays. This is cost recovery for a nonprofit research tool, not a commercial charge.
Open CR-Lite →
Background reading: CR Standard · RN010 — The Shape of No · RN011 — The Conditional Answer
⚖️
MMDE — Multi-Model Deliberation Engine
Live tool · Patent pending · Five AI providers · Quantum tier integration
What it does
Structured multi-model deliberation
Submit a research question across up to five frontier AI models simultaneously — Claude, GPT-4o, Gemini, Grok, DeepSeek — with domain intelligence templates, red-team modes, and quantum computing tier augmentation.
Who it's for
AI governance researchers, academics, policy analysts
Anyone studying how frontier AI systems converge or diverge on complex governance questions. RN011 was produced using the MMDE — the session IDs are published for independent verification.
Access
Suggested donation per session
Multi-model sessions call up to five paid APIs simultaneously. The suggested donation covers those direct costs. As a nonprofit, the Foundation cannot absorb unlimited free API usage.
Open MMDE →
Background reading: RN011 — The Conditional Answer · RN011A — Epistemic Movement