Risk Engine
Routes risk in proportion to impact — high-risk actions and low-trust counterparties escalate to a human, they don't fail silently.
What it does
The Risk Engine routes risk in proportion to impact. Policy atoms carry thresholds — a high-value action, an unverified counterparty, or a Trust Index below a review band — and instead of failing silently, they escalate to a human. Trust is used for guidance, never as a hard gate: a low score routes an action for review, it does not block it outright. When something does go wrong, incidents and near misses are recorded and classified for reportability with the reasons stated — legal or physical harm is reportable however the event turned out, and a foreseeable near miss that would have been serious is reported too, so an authority can see a hazard before it lands rather than after. Classification fails open: an unrecognised event is routed rather than quietly filed as noise.
Inside Risk Engine
The core functions this engine provides across the Trust Fabric.
Risk Engine composes with the Trust Fabric
Every engine is independently useful and interoperates through one API-first Trust Fabric — so identity, evidence, policy, risk and certification stay consistent across your AI estate.
Related engines
Registry Services
Authoritative, federatable on-chain registries for agent identities and published trust records.
- HCS-2 identity registry
- Agent DID registry
- HCS-28 trust registry
- Federatable / discoverable
Certification Authority
Issues living certifications backed by assurance-graded verification, continuous evidence and the Trust Index.
- Assurance-graded KYC/KYB
- Living certificates
- Digital trust marks
- Auto-revoke on drift
AI Sandbox
A public, no-KYB environment that hands you a governed agent — first allow / observe / block / escalate in minutes.
- Public no-KYB endpoint
- Shared, pre-issued test agent
- Live allow / observe / block / escalate
- Same code to production
See Risk Engine in action
Book a demo and see how the platform establishes trust across your AI systems.
