Compliance & Audit

Audit Ledger

Append-only. Agent-isolated. Immutable. Every action, permanently recorded.

Overview

The SinzAI Audit Ledger is the foundation of the platform's accountability guarantees. It records every agent action — reads, writes, API calls, data accesses, emails, and financial transactions — in an append-only, cryptographically chained store that no agent credential can reach. It is the single source of truth for compliance audits, incident investigations, and rollback operations.

Design Principles

📝Append-Only

Ledger entries can only be created, never modified or deleted. Any attempt to alter an existing entry is cryptographically detectable — the hash chain breaks.

🔒Agent-Isolated

No agent credential has write or delete access to the ledger store. The ledger lives in a separate security domain that only the SinzAI governance engine can write to.

⛓ī¸Cryptographically Chained

Each entry contains a SHA-256 hash of the previous entry. Tampering with any entry invalidates the entire chain from that point forward — instantly detectable.

🔍Complete Context

Every entry captures the full decision context: agent identity, model version, prompt fingerprint, data sources consulted, and the before/after state of any modified records.

↩ī¸Tamper-Proof Rollback

Rollback operations are themselves recorded as new ledger entries. You can audit not just what the agent did, but what the security team did to reverse it.

🕐Server-Side Timestamps

Timestamps are assigned by the server, not the agent. An agent cannot backdate an action or manipulate the temporal record.

Entry Structure

Each ledger entry contains 15 fields that together provide complete traceability for every agent action:

{
  "entry_id": "led_01J3XK7N2P8Q5V9W4R6A",
  "agent_id": "agent-001",
  "agent_did": "did:sinzai:agent:8f3a1b2c4d5e6f7a",
  "model_version": "gpt-4o@sha256:a1b2c3d4",
  "prompt_fingerprint": "fp_sha256:e5f6a7b8c9d0",
  "data_sources": [
    "ds_sales_leads_2026_us_east"
  ],
  "action_type": "email_send",
  "target": "hubspot:contacts:batch_send",
  "before_state": null,
  "after_state": {
    "emails_sent": 12,
    "recipient_ids": ["c_001"..."c_012"]
  },
  "rollback_op": "hubspot:email:retract",
  "server_timestamp": "2026-08-05T14:32:17.421Z",
  "previous_hash": "sha256:9a8b7c6d5e4f3a2b1c0d9e8f7a6b5c4d",
  "entry_hash": "sha256:1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6e",
  "trust_score_snapshot": 92
}

The rollback_op is pre-computed at write time — rollback is instant because the reversal operation is already known.

Cryptographic Chain Verification

The integrity of the ledger can be verified at any time by walking the hash chain. Each entry's previous_hash must match the entry_hash of the preceding entry. If any entry has been modified, deleted, or inserted out of order, the chain breaks at that point and the tampering is immediately detectable.

SinzAI performs automatic chain verification every 60 seconds and on every audit report generation. Any break triggers an immediate security alert with the affected entry IDs and the nature of the anomaly.

Query & Export

The ledger supports structured querying from the SinzAI console and API:

  • Filter by agent, action type, time range, data source, or trust score
  • Full-text search across prompt fingerprints and action metadata
  • Export to JSON, CSV, or PDF for external audit tools
  • Streaming API for SIEM integration (Splunk, Datadog, Elastic)
  • Chain verification report with cryptographic proof of integrity