Verification Playground
This is a live protocol demonstration. Submit a question — Piea retrieves real evidence artefacts, hashes every source with SHA-256, streams a grounded answer, and returns a validation result you can inspect. Questions outside AIEP's scope are rejected — the playground fails closed.
Evidence retrieval SHA-256 hashing Schema validation Hash commitment Replayable response Fails closed
Input Ask anything about AIEP — evidence is retrieved live
Try a sample question:
AIEP Response Evidence-grounded answer — generated only after evidence is retrieved and hashed
Run a verification above to see a live, evidence-grounded response.
Evidence Rail Every source retrieved — URL, SHA-256 hash, confidence tier, retrieval timestamp
Evidence artefacts will appear here after verification.
Validation Result Schema conformance + hash commitment — independently verifiable by any third party
Validation details will appear after verification runs.
What this demo proves
Every response above exercises the full AIEP protocol stack. Here is what each panel shows and why it matters for enterprise AI.
Evidence Rail
Real sources. Real hashes.
Every artefact in the Evidence Rail was retrieved live. Each carries a SHA-256 hash of the source at retrieval time. If the source changes, any replay will detect the discrepancy — the hash will not match.
Hash Commitment
Tamper-evident by construction.
The response hash in the Validation panel commits both the answer text and the evidence set. Any third party can independently verify the response has not changed since it was generated — without contacting Piea.
Schema Validation
Machine-checkable conformance.
The validation checks shown — schema conformance, hash presence, evidence rail structure — are the same checks any AIEP validator runs. The AIEP canonical schema is published openly. Any system can validate any AIEP response.
Fails Closed
Out-of-scope queries are rejected.
Try asking a non-AIEP question. The playground rejects it rather than fabricating an evidence-grounded answer. This is the AIEP fail-closed principle — the system does not degrade gracefully into hallucination.
Replayable
Deterministic. Reproducible.
The same question, same evidence, same protocol version will produce byte-identical hashes across any two conformant AIEP implementations. This is what makes AIEP suitable for regulated environments — outputs are reproducible from artefacts alone.
Confidence Tier
Source quality, not just source count.
The confidence tier shown in each evidence artefact reflects AODSR's five-class source taxonomy — from Tier 1 primary sources (legislation, official records) to Tier 5 community content. The response confidence ceiling is enforced by the lowest-quality source cited.
Use AIEP in your system
All components are open source under Apache 2.0. No registration required.
AIEP-VALIDATOR ↗
Validate AI outputs against AIEP schemas, evidence, and hashes
AIEP-MIRROR ↗
Publish machine-readable verifiable sources at /.well-known/aiep/
AIEP-GENOME-SDK ↗
Integrate AIEP into applications — frozen kernel R1–R8, Python, Apache 2.0
AIEP-VECTORS ↗
Canonical test vectors for deterministic validation and interoperability
Specification →
Normative protocol document — R1–R8 primitives, artefact types, conformance rules
Build with AIEP →
Full map of all public repositories and how they connect
See also: RAG vs AIEP · AI Audit Trail · Evidence Rail Example · Publish a Mirror · Machine endpoint · GitHub ↗