Aligned with AARM v1.0. Not certified against it.
A neutral specification is more useful to you than our marketing. This brief states which AARM v1.0 requirements our self-attestation records as satisfied, which one is still a gap, which is out of scope, and where the diffable record lives.
What AARM is
AARM — Autonomous Action Runtime Management — is the Cloud Security Alliance’s specification for governing autonomous AI agents at runtime. It was authored at Vanta, donated to the CSAI Foundation in April 2026, and is governed by a working group spanning Microsoft, Vanta, Noma Security, Zenity, Elastic, Truist and Darktrace. It gives a compliance owner a neutral yardstick to hold a vendor to, instead of the vendor’s own marketing.
Our claim, stated precisely
We publish “aligned with AARM v1.0” — not a certification. AARM has no certification body and we hold no attestation from a third party against it. What exists is our own self-attestation, published as a document you can diff.
Our current self-attestation records R1–R6 as satisfied at the Core level and R8 as satisfied, with R7 still a gap and R9 out of scope for our architecture. That document is the record, and it outranks this brief.
Status by requirement
The five decision verbs
AARM R4 defines five decisions. All five ship today: ALLOW, DENY, MODIFY, STEP_UP, DEFER. The ruling is returned by deterministic policy over a canonical input — no AI model makes the enforcement decision — and each governed outcome produces a product-specific decision record.
Four questions to ask any vendor, including us
2 What does policy see at evaluation time, and is the decision deterministic?
3 What decision evidence is produced, and what is required to re-evaluate it later?
4 Which requirements are claimed as satisfied, which are gaps, and where is that written down?