Compliance Enforcer
operationalThe Compliance Enforcer is the governance layer of the NANDA network — a policy decision engine that evaluates every agent interaction against configurable rules for PII protection, regional routing, capability restrictions, and regulatory compliance.
Core Capabilities
Policy Decision Point
Versioned policy documents define allow/deny rules. Each incoming agent interaction is
evaluated against the active policy set, producing an allow or deny decision with reasons.
PII Redaction
Automatic detection and redaction of personally identifiable information in agent payloads — enforcing data minimization principles before cross-agent communication.
Regional Routing
Jurisdiction-aware routing rules ensure data residency compliance. Agents in specific jurisdictions are routed to appropriate regional endpoints.
Violation Tracking
Every policy violation is recorded with the offending agent, envelope hash, reason, and timestamp — providing a complete compliance audit trail.
Decision Flow
When an agent interaction is submitted for compliance evaluation:
- Envelope received — the interaction payload (from agent, to agent, capability, data) is captured
- Policy evaluation — active policies are applied in version order against the envelope
- Decision recorded — an
allowordenydecision is stored with the full reasoning chain - Violation flagged — if denied, a violation record is created for audit and potential enforcement action
Policy Structure
Policies are stored as versioned JSON rule sets. Each policy contains:
policy_id Unique identifier for the policy
rules_json JSON-encoded rule set defining conditions, actions, and enforcement levels
version Integer version — higher versions take precedence during evaluation
A2A Protocol Actions
The Compliance Enforcer is accessible via the A2A JSON-RPC protocol at /a2a:
policy.eval Evaluate an interaction envelope against the active policy — returns decision, reasons, and optional PII-redacted envelope
policy.rules.get Retrieve the current policy rules, optionally filtered by policy ID
policy.violation Report a policy violation for an agent with reason and severity
Zero Trust Agentic Access (ZTAA)
The Compliance Enforcer implements the ZTAA framework — applying Zero Trust principles to AI agent interactions:
- Verify Explicitly — every interaction requires cryptographic verification, never implicit trust
- Least Privilege — agents are granted only the minimum capabilities needed per interaction
- Assume Breach — continuous monitoring and revocation ensure compromised agents are detected and isolated
Integration with Other Services
- Agent Registry — reads agent jurisdiction and capability data for policy evaluation
- Capability Certifier — compliance attestations may reference certification grades
- Observer Evaluator — compliance violations feed into reputation scoring
- Points Auditor — violation tracking integrates with the audit trail