Skip to main content
AI Forensics

Reconstruct. Verify. Prove It.

AI Forensics reconstructs an agent's entire decision chain, commits to it with a KMS-signed DSSE attestation, and exports evidence any auditor can verify offline — without trusting our infrastructure.

Monitoring tells you something broke.

Forensics tells you which agent did it, what it was trying to do, and whether the audit chain was tampered with.

Monitoring / Observability

  • What's happening now?
  • Metrics, traces, APM dashboards
  • Application-level logs (mutable)
  • Alert when something goes wrong

AI Forensics

  • What exactly happened, provably?
  • Cryptographic chain of evidence
  • Tamper-evident, independently verifiable
  • Reconstruct and replay any incident

Forensic Capabilities

Every capability is built into the platform — not bolted on after the fact.

Chain-of-Thought Capture

Every gateway decision — ALLOW, DENY, policy match, upstream call — becomes an HMAC-chained entry tied to the specific agent, policy version, and credential used.

Incident Replay

Given any agent and time window, reconstruct every request, every policy evaluation, and every outcome in order. Step through the full session timeline.

Signed Session Attestations

Every session closes with a DSSE envelope signed by a KMS-held ECDSA P-256 key. The signed payload commits to the exact audit range, so an auditor can prove offline that the chain hasn't been altered since sign time.

Chain Verification

One API call (or the offline CLI) verifies the integrity of your entire audit chain. If a single record was altered or deleted, the chain breaks — and we tell you exactly where.

Anomaly Detection

Automated detection of latency spikes, cost outliers, and deny clusters. Surface suspicious agent behavior before it becomes an incident.

Forensic Export

Export forensic reports as JSON with chain-of-custody certificates, or as CSV for spreadsheet analysis. Built for auditors, not just engineers.

Shadow Agent Detection

Automatically detect unregistered agents attempting to access your infrastructure. Surface rogue agents through denied request patterns.

The Four Pillars of AI Agent Governance

Most solutions cover one or two. AI Identity covers all four.

PillarCore QuestionAI Identity Capability
IdentityWho is this agent?Per-agent API keys, lifecycle management, scoped credentials.
PolicyWhat is it allowed to do?Fail-closed gateway, deny-by-default policy evaluation.
ComplianceCan we prove rules were followed?DSSE-signed session attestations, HMAC-verifiable audit logs, automated compliance assessments.
ForensicsWhat happened, provably?Hash-chained logs, incident replay, export, offline cryptographic verification.

Forensics Meets Compliance

AI Forensics isn't optional — it's what regulators are already requiring.

EU AI Act

Traceability & record-keeping (Art. 12)

HMAC-chained audit trail with full request metadata, configurable retention, and evidence export.

SOC 2 Type II

Tamper-evident audit logging

DSSE-signed session attestations (ECDSA P-256) over an HMAC-SHA256 hash chain. Dual-key separation: signing keys in KMS hardware, chain HMAC keys in the application tier — forging evidence requires both.

HIPAA

Audit controls (164.312(b))

Per-agent activity logging with attribution, chain-of-custody certificates for evidence export.

NIST AI RMF

Agent observability & integrity

Continuous monitoring, anomaly detection, forensic replay of agent decision sequences.

If it didn't go through AI Identity, you can't prove what your agent did.

Start with the free tier — 5 agents, tamper-proof audit trails, and chain verification included.