CSOAI OS · the ontology
The semantic layer for AI governance.
What Palantir's Ontology did for enterprise data — turning scattered records into objects, links and actions you can reason over — this Council Ontology proposes the same pattern for AI governance. It catalogues representative object types and relationships; it does not claim every framework, organisation or agent is covered or governed. Toggle the Ontology layer to explore the visual map.
Objects — what the world is made of
◈
AI systems & agents
every model, agent and humanoid under governance
▤
Frameworks & law
EU AI Act, NIST, ISO 42001, GDPR, cyber (CRA/NIS2/DORA)…
▦
Organisations
Fortune 500/100, suppliers, public bodies
❖
Governments & regulators
AI authorities and jurisdictions
◎
Threats & incidents
rogue swarms, deepfakes, cyber events
🛡
People
executives, officials, creators — and everyone
Links — how they relate
governs
a government / framework governs an AI system
applies-to
a framework applies to an organisation or sector
defends
a cyber control defends an asset against a threat
monitors
the Watchdog monitors objects for incidents
crosswalks
one control satisfies many frameworks at once
signs
a specifically published record can carry a verifiable signature; coverage is record-specific
Actions — what you can do on any object
Govern →
open a classification or assessment workflow and inspect its evidence state.
Simulate →
run the outcome across thousands of branches before acting.
Stop →
explore a containment design; no live intervention is claimed.
Prove →
open the System Card workflow and verify any artifact it publishes.
Crosswalk →
inspect published mappings without treating them as a compliance determination.
Protect →
shield a person's identity and likeness from deepfakes.
One proposed ontology, with open object and link mappings. Any signed evidence remains artifact-specific; ordinary actions are not automatically governed or signed. Explore the catalogue from one place.
