About AIEP
What is AIEP?
The Architected Instruction & Evidence Protocol (AIEP) is an open protocol that lets organisations publish structured, verifiable artefacts — documents, records, certificates, schemas — in a way that both humans and AI systems can retrieve, validate, and trust.
Think of it as a publishing standard for the AI era. Instead of a webpage that a model must interpret and guess at, an AIEP-compliant publisher provides machine-readable files at a known location, conformant to a canonical schema, with an unbroken chain from the artefact back to its issuer.
Why does this exist now?
AI models are increasingly used as a replacement for search. People ask a model a question and expect a reliable answer — often about a company, a product, a legal claim, or a technical specification.
The problem: models retrieve content from the web, but the web was not built for machine verification. A webpage can say anything. There is no standard way for a model to check whether a claim is current, who made it, or whether it has been tampered with.
AIEP solves this by giving publishers a structured, verifiable surface — the Mirror — that AI systems can query with confidence.
What does “verifiable” mean here?
An artefact is verifiable when:
- it is published at a known, stable path (
/.well-known/aiep/) - it conforms to a canonical schema so its structure is machine-checkable
- it carries a hash so a retriever can confirm the content has not changed
- it is linked to an issuer identity in the AIEP registry
A model retrieving a verified artefact can confirm: this came from this publisher, it matches this schema, and it has not been altered since publication.
Who is AIEP for?
| Role | How AIEP helps |
|---|---|
| Publishers | Structured surface for AI retrieval; reduces hallucination risk about your content |
| AI systems / agents | Verifiable, schema-conformant data source; no guesswork |
| Developers | Open protocol, no licence fee, self-hostable |
| Auditors / regulators | Immutable evidence trail via the Innovation Ledger |
Open use
AIEP is an open protocol. Anyone can implement it. Certification exists only to protect the meaning of “AIEP Certified” — it is not required to use or publish under the protocol.
See: Protocol · Quickstart · Mirror
It is a protocol that links instructions to evidence, enabling the web to become a verifiable knowledge infrastructure for humans and artificial intelligence.
AIEP is built around a simple premise: if information is going to be relied upon—by people, by organisations, or by machines—then it should be possible to see what supports it. A claim without evidence is not knowledge. It is noise.
AIEP does not attempt to “win” the internet with a single model or a single platform. Instead it makes a different bet: that the web can evolve into a distributed knowledge layer, where publishers expose machine-readable artefacts that can be retrieved and validated from the source.
Dissent and outliers
Real discovery is not a smooth consensus curve. It is disagreement, conflict, and eventually evidence. Many ideas that are mocked today become reasonable tomorrow when new artefacts appear. AIEP preserves this reality by making it normal to store:
- consensus
- outliers
- radical outliers
and to recall old ideas when new evidence makes them relevant again.
Mirror: the dissemination layer
AIEP Mirror is how this becomes real on the open web. Sites publish /.well-known/aiep/ endpoints that allow automated systems to discover, retrieve, and validate artefacts.
In practice this means the web can become less about searching pages and more about retrieving knowledge backed by evidence from where it originates.
Piea — the architecture in production
Piea is an evidence intelligence layer that plugs into existing SaaS products, databases, and applications — adding verifiable audit trails, hash-bound outputs, and governed retrieval without a rebuild. It implements the complete AIEP Piea Surface (P116–P128): GENOME R1–R8 cryptographic primitives, live evidence retrieval with source integrity inspection, Dissent Signal Engine, Replayable Reasoning Chain, Semantic Branch Detection, multimodal document ingestion, and governed file output. Chat is one interface; integration is the product.
Piea is not a demo. It is the proof that AIEP is complete and coherent enough to build production systems on.
AIEP Miner — governed compute research layer
AIEP Miner is a private infrastructure research project exploring how AIEP-governed workloads may operate across constrained, repurposed, or distributed compute environments. It investigates evidence hashing, validation, replay, node attestation, and reasoning-support tasks under governed execution constraints.
AIEP Miner is not publicly disclosed. A controlled headline page is available at aiep.dev/aiep-miner. NDA-gated briefings are available on request.
AIEP Miner — controlled disclosure →
AI-assisted drafting
AI tools have been used to assist in generating documentation and code during development. The ideas, principles, and overall thesis behind AIEP originate with Neil Grassby.
Knowledge grows when shared.