AIEP

Verifiable AI answers, backed by real evidence

AIEP produces answers using real sources, validates them, and binds every output to cryptographic proof you can check.

Explore Hub →    ⭐ View on GitHub →

Try it in 60 seconds

pip install aiep-genome-sdk

from aiep.genome import verify

result = verify("Is inflation rising?")
print(result)

Output:

{
  "answer": "...",
  "sources": ["https://...", "https://..."],
  "hash": "0xabc123...",
  "confidence": "verified"
}

See the difference

Standard AIAIEP
Generates answersProduces evidence-bound outputs
No guaranteed sourcesReal, retrievable sources
No audit trailCryptographic proof
Non-deterministicDeterministic & replayable

Why AIEP exists

Most AI systems generate answers by prediction.

  • They can be confident but wrong
  • They rarely provide verifiable sources
  • There is no reliable audit trail

AIEP changes that by requiring evidence before output.

What AIEP is

AIEP (Architected Instruction & Evidence Protocol) is a protocol for producing evidence-bound outputs.

It is not a model.

It defines how systems:

  • retrieve data
  • validate sources
  • bind outputs to evidence
  • enforce deterministic behaviour

AIEP inference

AIEP does not simply attach sources to answers.

It governs how evidence is retrieved, validated, weighted, and used to produce outputs.

Each response is:

  • evidence-bound
  • confidence-scored
  • reproducible
  • fail-closed where admissible evidence is missing

AIEP guarantees

Every governed response:

  • includes an Evidence Rail
  • references real source artefacts
  • carries a confidence tier
  • is deterministic and replayable
  • is fail-closed if evidence is insufficient
  • is machine-discoverable via /.well-known/aiep/

Core components

  • Hub — developer gateway
  • Piea — AIEP reasoning engine
  • Mirror — machine-readable web layer
  • GENOME SDK — integration toolkit

What you can build

  • AI systems that prove their outputs
  • Applications with audit-grade reasoning
  • Workflows requiring regulatory alignment
  • Systems that do not silently hallucinate

Machine-readable by design

/.well-known/aiep/
/.well-known/aiep/schema

These allow:

  • AI agents to discover capabilities
  • Systems to validate evidence structures
  • Cross-platform interoperability

Direct endpoints: /.well-known/aiep/index.json · /.well-known/aiep/schema · /.well-known/aiep/metadata.json

Start building

Explore Hub →    View GitHub →    Get Started →