Why Now
Three forces are converging simultaneously. Each one alone would be insufficient. Together they make AIEP not just timely but necessary.
1. The regulatory wave has arrived
Governments and regulators have spent years discussing AI governance. In 2024 and 2025, enforcement began.
| Regulation | Key date | What it requires |
|---|---|---|
| AI Act | Enforcement: Feb 2025 | Risk classification, transparency obligations, human oversight, audit trails for high-risk AI systems |
| AI Liability Directive | Expected: 2026 | Causal attribution of AI-caused harm — requires traceable evidence chains |
| GDPR Article 22 | In force | Explainability of automated decisions affecting individuals |
| AI Safety Framework | 2024 | Evaluation and audit requirements for frontier and high-risk AI |
| FDA AI/ML SaMD Guidance | 2024 | Predetermined change control plans and audit trails for AI medical devices |
| FCA Model Risk Management | 2024 | Governance and audit trails for AI used in regulated financial services |
| ISO/IEC 42001 | 2023 | AI management system standard — documentation, traceability, and governance |
| MAS Model Risk Guidelines | 2024 | Model governance and audit requirements for AI in financial services |
Every one of these requires the same thing: a machine-readable, tamper-evident, auditable record of AI decisions and their evidence basis.
AIEP produces those records as a structural property of every artefact — not as a compliance add-on constructed after the fact.
2. AI deployment has outpaced AI evidence
The current generation of AI systems — large language models and retrieval-augmented systems — has a structural limitation that cannot be fixed by scaling:
| Limitation | What it means in regulated deployments |
|---|---|
| Training cutoff | A model cannot know what changed after its training date without live retrieval |
| Unverified retrieval | RAG systems retrieve web pages — there is no protocol signal that the content is current, accurate, or evidenced |
| Hallucination | When verified retrieval fails, models generate plausible text — there is no mechanism for a consumer to distinguish which |
| Invisible provenance | A model cannot tell a regulated enterprise which document it used, when it was published, who issued it, or whether it has changed |
| No execution gate | Current retrieval systems have no mechanism to block action on a claim that lacks sufficient evidence |
These are not bugs in any particular model. They are architectural properties of the current web — which was not designed for machine-verifiable knowledge retrieval. AIEP redesigns the evidence layer without requiring any change to existing web infrastructure.
3. The evidence layer has never existed until now
The web has had every other fundamental layer for decades. The evidence layer has been conspicuously absent:
| Layer | Protocol | Established |
|---|---|---|
| Document | HTTP / HTML | 1991 |
| Linking | URL / Hypertext | 1991 |
| Structured data | XML / JSON | 1998 / 2001 |
| Semantic metadata | schema.org / JSON-LD | 2011 |
| Machine-readable APIs | REST / OpenAPI | ~2000s |
| Evidence | AIEP | 2025 |
The absence of an evidence layer was not a priority when the web was primarily human-readable. It is a critical gap when AI systems use the web as a knowledge substrate for high-stakes decisions.
4. The patent window is open
9 foundational patents covering the core AIEP stack were registered in November 2025. A further 9 constitutional stack applications were filed on 07 April 2026, and a further 6 AGI cognitive architecture applications on 08 April 2026 — 24 applications total, all with confirmed GB numbers. Each batch carries its own 12-month priority window for international filings (PCT, EPO, USPTO): the November 2025 applications’ window closes November 2026; the April 2026 applications’ window runs to April 2027.
After that window, the prior art is public. The protocol remains open and adoptable, but the IP protection window has closed.
For organisations that want to operate ahead of the commodity phase — building on the protocol while the IP position is still active — the window is now.
The convergence
| Force | Current status |
|---|---|
| Regulatory demand for AI audit trails | Active — AI Act enforcement started February 2025, FCA and FDA requirements in force |
| AI deployed without an evidence layer | Accelerating — every major enterprise is deploying AI into regulated contexts without this |
| Evidence layer protocol | Ready — AIEP is fully specified, open source, and testable today |
| International patent protection window | Open — closing November 2026 |
The evidence layer problem was always going to need solving. The regulatory wave means it needs solving now. AIEP is the solution that exists and is available today.
The story at launch
Piea is proof that AIEP solves the bloated LLM problem. The mirrored web is the brain. Piea is the interface. Every claim is hash-bound to a specific source at a specific moment. No training data. No knowledge cutoff. No hallucination on source-resolved queries. The only AI substrate with filed patents covering hardware-level governance attestation and goal-trigger enforcement. The architecture already supports it. The build makes it visibly, demonstrably, verifiably true at launch.
Related
- Vision — what AIEP is building toward
- Regulatory Governance — how AIEP satisfies specific regulatory requirements
- Access Tiers — how to engage with AIEP now
- Explainer — one-page summary
- Quickstart — start building today
- Contact — enterprise and regulatory enquiries