For compliance and risk
Which framework maps to which evidence, what a report contains, what it does not claim, and what we do not map at all.
The evidence underneath everything.
Every policy decision is written as an audit event in a SHA-256 hash chain, with hourly Merkle root snapshots and a chain-verification call you can run yourself over any date range. Deletion and modification after the fact are detectable. Events carry the agent, the tool, the arguments, the decision and the reason, with PII redaction applied on the way in, and they export to your own systems.
What maps to what.
- SOC 2 — seven Trust Services Criteria (CC6.1, CC6.2, CC6.3, CC7.1, CC7.2, CC7.3, CC8.1), with the access-control and change-monitoring criteria evidenced from live proxy decisions rather than from screenshots.
- ISO/IEC 42001 — eight Annex A controls are machine-attestable from live evidence (A.2.2, A.6.2.6, A.6.2.8, A.7.5, A.8.4, A.9.2, A.9.4, A.10.3). The rest of Annex A and the management-system clauses are document evidence.
- EU AI Act — Annex IV technical documentation, and Article 73 serious-incident reporting. High-risk obligations apply from 2 December 2027 for Annex III systems and from 2 August 2028 for Annex I products, under Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744; the prohibited-practice articles already apply.
- NIST AI RMF 1.0 — eight subcategories are machine-attestable from live evidence. The rest are document evidence, and the report says which is which rather than blurring the two.
What the report is not.
It is not a SOC 2 Type I or Type II opinion, it is not issued by an accredited body, and it is not a certificate of conformity. It is the evidence pack an auditor or a customer’s security team asks for. Every generated report carries that statement on its own face — we did not put it in the small print of a marketing page.
What we do not map, and will not pretend to.
There is no DORA mapping, no NIS2 mapping and no ENS mapping in the product today. If you are subject to those, what we give you is the underlying record — which agent, which tool, which data, what was blocked and why — which is the material those obligations are answered with. Anyone who offers you a finished DORA report generated from agent traffic is describing something that does not exist yet.
Start with a pilot
Two weeks, in Shadow Mode, against your own agents. At the end you get the call log, the policy set we would have enforced, and the list of calls that would have been blocked. If that list comes back empty, that is a useful answer too.