Decifact

Two systems can both pass audit. That doesn't mean their decisions can be reconciled.

Decifact checks whether two independently governed decisions were ever on the same basis to begin with — before anything depends on it.

Read the Quickstart → Explore Decifact →

Both hospitals followed their own governance process. Here's what happens when their decisions need to work together:

Hospital A

UK-AI-Framework-v1
"deploy_ai_triage approved"
FORMALLY_INCOMPARABLE
comparability result

Hospital B

UAE-AI-Framework-v2026
"deploy_ai_triage approved"

The hospitals are not disagreeing.

They are operating under different governance references.

Scenarios

When valid is not enough

Hospital Triage

Hospital A (UK framework) and Hospital B (UAE framework) both approve AI-assisted triage. Same intent. Different governance foundations. Are they comparable?

Cross-Ministry

Agriculture Ministry and Finance Ministry both approve the same AI system. Each approval is internally valid. Can they be placed on the same comparison basis?

Multi-Agent

Agent A produces: BUY NVDA
Agent B produces: SELL NVDA
Both operating under valid policies. Is this a disagreement — or an incomparability?

Classifications

Three outcomes

EQUIVALENT

Shared comparison basis exists. Governance conditions are identical.

NON_EQUIVALENT

Shared comparison basis exists. Conditions differ. Valid disagreement.

FORMALLY_INCOMPARABLE

No shared comparison basis detected. Not a disagreement. A structural finding.

FORMALLY_INCOMPARABLE is not an error. It is a first-class governance result.

Decifact

Decifact does not govern either system.

Comparability

What comparability cannot determine.

  • Whether either decision is correct — two decisions can be comparable and both wrong
  • Whether a system's behavior conforms to its own declarations — that's a claim-vs-conduct question, not a claim-vs-claim question
  • Whether a comparability result remains stable as either system evolves — that's a separate question from comparability at the time it was evaluated
  • Whether a decision or its governance conditions can be independently reconstructed later
  • Whether a decision should be executed — a decision can be comparable and still be inadmissible

Start here

QUICKSTART.md — understand and run Decifact in 5 minutes
Playground — view source, run locally with FastAPI
Requires local API server. See QUICKSTART.
API Reference — /compare /canonicalize /verify-equivalence

Working on cross-system decisions?

Building multi-agent systems, cross-domain governance, or decision verification infrastructure?

If what you saw on this page maps to a problem you're working on — let's talk.

xsa52000@gmail.com