Reference updated 2026-05-21

Compliance mapping

How Cullis Mastio primitives map to specific clauses in EU AI Act, DORA, EIOPA, IDD, NIST AI RMF, Colorado AI Act, ISO 42001, HIPAA, and other frameworks regulated deployments encounter.

Compliance mapping

Compliance teams need to map Cullis capabilities to specific clauses in their framework. The table below is the working map for the current Mastio release. It is a reference, not a certification: a specific compliance assessment for any regulated deployment remains the responsibility of the deploying organization.

FrameworkArticle / clauseCullis capability
EU AI ActArt. 12 (logging), 13 (transparency), 14 (human oversight)Tamper-evident audit chain, per-decision reasoning capture, ADR-020 4-quadrant identity
EU AI Act Annex IIIHigh-risk classification (credit, insurance pricing, recruitment)Per-agent identity + per-decision audit + 4-eye approval capability
DORAArt. 28 (third-party ICT)mTLS + DPoP cryptographic identity per action, self-hosted deployer model
EIOPAAug 2025 Opinion 8-axis governanceMastio dashboard + audit export + policy decision logging across fairness, data, record, transparency, oversight, accuracy, robustness, cyber
IDDArt. 17 (fair treatment), Art. 20, 30Per-claim audit chain, exportable signed bundle for customer complaint reconstruction
EU GMP Annex 22AI/ML in pharma manufacturingPer-agent identity + immutable change log + signed releases for validated lifecycle
NIST AI RMFMEASURE 2.7, GOVERN 1.7Standardized audit log export, per-agent identity, role separation
Colorado AI ActConsumer disclosure for high-risk AI (June 2026)Decision logging with reason, per-decision provenance
ISO 42001AI Management System controlsOperational governance, lifecycle controls, audit evidence
HIPAA164.312(b) audit controlsAppend-only audit chain, integrity controls, external verification

How to read this table

Each row pairs a regulator’s requirement with the Cullis Mastio primitive that addresses it. Most rows reduce to one of three Mastio building blocks: per-agent x509 + SPIFFE identity, the policy decision point that runs before every LLM or MCP call, and the hash-chained append-only audit log.

The mapping is conservative. Where a clause requires evidence Cullis cannot itself produce (e.g. data-quality management for the training set under EU AI Act Art. 10), the row is omitted rather than stretched.

Further reading

The research that backs this mapping (long-form articles on EU AI Act Annex III, DORA, EIOPA, and the broader regulated-AI landscape) is published at mazzolad.com/writing.