When AI decisions are questioned,
explain what happened
Works with partial logs and manual oversight records. No production integration required to start. Provides evidence and analysis support. Does not determine compliance or liability.
Decision Path Reconstruction
Observe how AI and human decision events are captured and reconstructed—not judged or scored.
Agent initiated $50,000 transfer request to external account.
Transfer amount flagged against automated limit threshold of $10,000.
Manual override recorded. Context: Emergency Liquidity Event.
Accountability path reconstructed for review. Chain integrity preserved.
This timeline illustrates forensic reconstruction capability. It does not make compliance determinations or assign liability.
Start With What You Have
We don't require perfect logging. We work with partial data and mark gaps as unknown.
No SDK or agent required
Upload what you have. We reconstruct the decision path and explicitly mark missing evidence.
Explicit Uncertainty
We surface what is known, what can be inferred, and what remains unknown. No guesswork. No false confidence.
Directly observed in evidence
Agent ID, timestamp, logged action
Derived deterministically
Time divergence, handoff detection, sequence analysis
Missing evidence and explicitly flagged
Human rationale not captured, partial logs
This transparency builds trust with regulators and internal reviewers who need to understand the limits of what can be reconstructed.
Forensic Event Feed
Every AI and human decision event captured with cryptographic proof. Immutable audit trail for regulatory investigation and review.
Important: AegisTrace provides evidence and analysis support. It does not determine compliance, violations, or liability. Demonstration visuals and data.
Decision Event Schema
Illustrative data structure for evidentiary capture. Demonstrates how events are cryptographically recorded for review.
Important: AegisTrace provides evidence and analysis support. It does not determine compliance, violations, or liability. Demonstration visuals and data.
{
"event_id": "evt_2026_001_au_nsw",
"timestamp": "2026-01-13T21:15:42.123Z",
"agent_id": "agent_legal_counsel_01",
"jurisdiction": "AU-NSW",
"regulatory_scope": [
"APRA CPS 230",
"Privacy Act 1988"
],
"event_type": "AI_EVENT_OBSERVED",
"decision_stage": "RECONSTRUCTION_COMPLETE",
"chain_hash": "sha256:a7f5c8d9e2b4f1a8c6d3e9b2f4a1c8d6e3b9f2a4c1d8e6b3f9a2c4d1e8b6f3a9",
"previous_hash": "sha256:b8f6d9e1a3c5f2b9d7e4f1a9c7d4e2b5f3a1c9d7e4b2f5a3c1d9e7b4f2a5c3d1",
"data_provenance": "CUSTOMER",
"hashed_reasoning": "sha256:c9g7e1b4d6f3a2c9e8b5f1d7a4c2e9b6f3d1a8c7e5b2f9d4a1c8e6b3f2d9a5c7",
"observed_action": "Document classification performed",
"human_action_present": false,
"policy_reference_ids": [
"APRA-CPS-230-3.1",
"PRIVACY-1988-5.2"
],
"policy_context_available": true
}Cryptographic fingerprint linking events in sequence
Reference to prior event creating unbroken chain
Category of observed event (AI action, human intervention, etc.)
Current reconstruction phase of the decision path
SHA-256 hash preserving AI deliberation for later review
Human-readable description of action recorded
Tracks origin of data (customer, third-party, hybrid)
Links to applicable policy contexts for investigative review