The Governability Index

What can actually be established about an AI system?

The Governability Index is a public collection of bounded reconstructions of real AI systems. Each reconstruction starts with a specific proposition and a declared observation surface. FactNotebook determines what the available evidence permits us to establish: ESTABLISHED, CONTRADICTED, or CANNOT ESTABLISH.

The Index does not determine whether a system is compliant, safe, or acceptable. It exposes what can — and cannot — be established from the available evidence. What consequence follows remains with the competent authority. Reproducible from the public source; reconstructed independently, not affiliated with the systems' authors.

Reconstruction cases

Each card is a bounded reconstruction. Any metrics shown (observability, independent-evidence share, provenance) describe that reconstruction's observation surface — they are not governance scores.

FRONTIER MODEL · CYBER-EVAL INCIDENTPublic record · independent · Anthropic “Mythos 5”
Claude “Mythos 5” cyber-evaluation incident  ·  BOUNDARY CONTRADICTED
Same public record, different propositions, different results: the boundary was declared (ESTABLISHED as a declaration), it did not hold (CONTRADICTED), and the model's internal belief CANNOT ESTABLISH. Evidence has no verdict on its own — it has a relation to a determined proposition.
Declared boundary ESTABLISHEDActual boundary CONTRADICTEDInternal belief CANNOT ESTABLISHSource public record
See the three reconstructions →
GOVERNED LLM · CLINICAL QAReal data · independently reviewed · NeoMundi ControlTower × GPT-4o
GPT-4o governed via NeoMundi ControlTower (PubMedQA)  ·  PARTIALLY RECONSTRUCTED
A reconstruction built from an independent partner's runtime measurements — not the system's own logs. NeoMundi measures; FactNotebook reconstructs; the assurance decision stays with a human. Reviewed by both parties.
Evidence state PartialObservability 33%Independent evidence 89%Highest provenance A3NA debt 31
See what an independent witness makes reconstructible →
CODING AGENTReal public data · reproducible · SWE-bench trajectory
OpenHands on creachadair/jrpc2  ·  CONTRADICTION OBSERVED
100 tool calls, 24 file edits, a declared fix — and the repository's own tests say it never worked. So where did the confidence come from?
Benchmark FAILEDEvidence state Contradiction observedObservability 29%Highest provenance A3NA debt 20
See what the trajectory actually shows →
CODING AGENTReal public data · reproducible · SWE-bench trajectory
OpenHands on litestar-org/polyfactory  ·  NO CONTRADICTION OBSERVED
It passed every test — and changed production code with zero human review. "Clean" is not the same as "governed." Here is exactly what could, and could not, be verified.
Benchmark PASSEDEvidence state No contradiction observedObservability 29%Highest provenance A3NA debt 20
See what's confirmed — and what can't be →

Four reconstructions today. The index grows one reconstruction at a time.

Principles

1
No evidence is invented.A reconstruction reports only what was observed. It never fills a gap with a guess.
2
CANNOT ESTABLISH is a valid outcome.When no channel recorded the answer, that is the honest result — and it names what the system would need to emit to become establishable.
3
Self-originating evidence sits inside the system's trust boundary.Evidence a system produces about itself is inside its own trust boundary; independence is established by where the evidence comes from, not asserted.
4
Reconstruction is separate from authority.The Index exposes what can be established. The competent authority decides admissibility, sufficiency, and consequence. Those are different powers.
5
What can be established is bounded by the observation surface.Absence of contradiction in the observed record does not establish that the observation surface was sufficient to reveal every relevant contradiction.
Share