Compliance

Where Ledgerline maps to existing regulation.

We are not a compliance product. We are a control plane that produces the evidence a compliance officer needs. This page maps Ledgerline's primitives to the regulatory frameworks that buyers most often ask about.

This is not legal advice. Use this mapping as a starting point for your own counsel, not a substitute for it. Final responsibility for compliance rests with the operator, not the platform.

Canada — AIDA (Artificial Intelligence and Data Act)

AIDA is expected to require operators of high-impact AI systems to maintain records of how the system is used, who is accountable, and what mitigations exist. The text continues to evolve. The shape of the obligation, though, is settled enough that we can map to it now.

AIDA-style obligationWhat Ledgerline provides
Identification of the natural person or organization responsible for an AI system's outputIdentity hierarchy ending at a human principal; every action carries the chain at decision time
Records sufficient for an audit of system useHash-chained, append-only audit log with policy and budget context per decision
Risk-mitigation controls before high-impact actionsPolicy engine + per-action budget reservation + kill switch
Notification of material harm or malfunctionAudit query API surfaces denied actions and chain-integrity breaks; webhooks on policy-violation events

European Union — AI Act

The EU AI Act applies risk-tiered obligations to "high-risk" AI systems. For autonomous agents that take consequential actions, several articles are directly relevant:

EU AI Act obligationWhat Ledgerline provides
Article 12 — Logging, automatic recording of eventsAppend-only, tamper-evident decision log with cryptographic integrity check
Article 13 — Transparency to deployersAuthorization-decision API returns reasons (which policy fired, which budget was hit) so the deployer's UI can surface them
Article 14 — Human oversightCo-signer policies, threshold-based escalation, and instant kill-switch revocation
Article 15 — Accuracy, robustness, cybersecurityHash-chained audit; policy and budget enforcement applied at every action

Canada — OSFI E-23 (Model Risk Management)

OSFI E-23 governs how federally regulated financial institutions manage model risk. AI agents that take financial actions are squarely in scope.

OSFI E-23 expectationWhat Ledgerline provides
Inventory of models and their useIdentity registry + policy attachments per identity; queryable inventory
Approval workflows before deploymentPolicies are versioned, attributable to a human approver, and required at decision time
Ongoing monitoring of model behaviourAudit log of every decision; anomaly review via export
Effective challenge / independent reviewAudit log is read-only and hash-chained — independent reviewers can verify it without access to live systems

United States — NYDFS Part 500

NY DFS Cybersecurity Regulation 23 NYCRR 500 applies to many financial services entities operating in New York. Several controls are directly satisfied by Ledgerline primitives.

23 NYCRR 500 controlWhat Ledgerline provides
500.06 — Audit trailHash-chained audit log retained per tenant policy
500.07 — Access privilegesIdentity revocation with cascade; periodic review queries
500.09 — Risk assessmentPer-identity policy and budget posture is queryable as a snapshot for risk-assessment exercises

SOC 2 trust service criteria

Ledgerline's design supports SOC 2 Type II audits both for the platform itself and for our customers' control attestations.

TSCHow Ledgerline supports it
CC6 — Logical access controlsIdentity hierarchy, revocation, privileged action audit
CC7 — System operations / monitoringAuthorization decisions and anomalies are surfaced via audit log and webhook
CC8 — Change managementPolicy changes are versioned; identity tree is journaled
P (Privacy)Audit log is per-tenant and segregated; data export and erasure supported through tenant API

Our own compliance posture


If your procurement team needs a specific regulatory citation we haven't covered, write to [email protected]. The fastest way to get coverage added here is to ask.