Framework / SOC 2

SOC 2 covers your systems.
Does it cover your agents?

A SOC 2 Type II tests whether you monitor for anomalies (CC7.2) and respond to incidents (CC7.3). Your tooling does that for infrastructure and people — AIR does it for autonomous agents, with signed evidence your auditor can verify.

SOC 2 Type IICC7.2CC7.3agent evidence

Book an agent audit See the evidence ↓

Continuous across the window · auditor-verifiable

SOC 2 evidence · live

CC7.2anomaly detected · 16 detectors

CC7.3evaluated & responded

Type IIsigned across the window

Agent actionhalted · SV-SCOPE

signed in-process · anchored Rekor

✓ auditor-verifiable · FRE 902(13)–(14)

01The stakes

SOC 2 Type II · observation window

Your SOC 2 covers the infrastructure. The agents acting on it slipped out of scope.

CC7.2 asks how you detect anomalies and CC7.3 how you respond. For servers and people you have answers; for autonomous agents, the control owner usually doesn't. AIR produces that evidence continuously, across the whole observation window.

02The criteria

Two criteria the agent layer keeps failing quietly.

CC7.2

detect

Monitor for anomalies and security events. AIR reads every agent action against 16 detectors (OWASP Agentic + LLM + AIR-native), continuously.

14 of the 16 run offline with zero config.

CC7.3

evaluate & respond

Evaluate and respond to security events. The causal graph and replay reconstruct what an agent did and why, so response is evidence-backed.

Containment can halt an agent before harm, signed.

Type II

over time

Type II tests controls across a window, not a snapshot. AIR signs each action in-process, so the evidence spans the whole period.

No backfilling the week before the report is due.

03How AIR answers

Every question a SOC 2 asks of the agent layer.

How do you detect anomalous agent behavior? (CC7.2)

Monitor16 detectors on every agent action, continuously.

How do you respond to an agent incident? (CC7.3)

InvestigateCausal graph and counterfactual replay reconstruct what happened.

Is the evidence continuous over the window?

ProveSigned in-process, every action, across the period.

Can the auditor verify it independently?

ProveAnchored to a public transparency log.

Mapped to the criteria, not raw logs?

ReportRecords map to CC7.2 / CC7.3.

04The evidence

Evidence mapped to the criterion, not a log dump.

A signed, timestamped record that names the criterion it satisfies — independently verifiable, and admissible under FRE 902(13)–(14) if the question ever leaves the audit room.

See all framework mappings →

SOC 2 evidence · agent layer

CC7.2 · anomaly detected — ASI02 tool misuse

CC7.3 · evaluated & responded — agent halted

Window · signed across the Type II period

signed in-process · blake3 · ed25519 · anchored Rekor

✓ auditor-verifiable · search.sigstore.dev

05What you get

The tiers SOC 2 teams choose.

Enterprise

most teams here

CC7.2 / CC7.3 agent-layer evidence
Causal graph, query & replay
Containment — halt agents before harm
SIEM: Splunk · Datadog · Sentinel · Sumo
SSO / OIDC, SLA

Book an agent audit

Air-gapped

regulated · sovereign

Everything in Enterprise, plus

On-prem / offline anchoring
No phone-home — records never leave
Extended retention + PQ re-anchoring
Admissibility Pack — FRE 902 + expert support

Talk to us

Bring the agent layer into your SOC 2.

A free agent audit shows exactly what CC7.2 / CC7.3 evidence you'd have for your agents today.

Book an agent audit →