CRP-SPEC-010: Regulatory Controls Mapping¶
Document: CRP-SPEC-010
Title: Context Relay Protocol (CRP) — Regulatory Controls Mapping
Version: 3.0.0
Status: Draft — PUBLIC DOCUMENT (intended for NIST/DISR submission)
Author: Constantinos Vidiniotis, AutoCyber AI Pty Ltd
Date: 2026-05-25
License: CC BY 4.0
Abstract¶
This document maps every CRP feature, header, and mechanism to specific controls in the EU AI Act, GDPR, ISO/IEC 42001:2023, NIST AI RMF 1.0, SOC 2 Type II, and the Australian AI Ethics Framework. It serves as both a protocol specification appendix and a standalone reference for regulators, auditors, and compliance teams evaluating CRP as a technical implementation mechanism for AI governance requirements.
1. EU AI Act (Regulation 2024/1689)¶
| Article | Requirement | CRP Feature | CRP Header | CRP Comply Output |
|---|---|---|---|---|
| Art. 5 | Prohibited AI practices | CRP-Compliance-EU-AI-Act: UNACCEPTABLE triggers HTTP 451 halt |
CRP-Compliance-EU-AI-Act |
Prohibition verification record |
| Art. 6 | Risk classification | Per-call risk classification based on registered AI system type | CRP-Compliance-EU-AI-Act |
Risk classification register |
| Art. 9(1) | Risk management system | DPE 13-stage pipeline as continuous risk assessment | CRP-Safety-Hallucination-Risk, CRP-Safety-Hallucination-Score |
Continuous risk assessment log |
| Art. 9(2)(a) | Identification of foreseeable risks | DPE identifies hallucination, fabrication, distortion, contradiction, omission risks per-call | CRP-Safety-Fabrications, CRP-Safety-Distortions, CRP-Safety-Omissions |
Per-call risk identification |
| Art. 9(4) | Residual risk mitigation | Safety Policy enforcement (halt-on, upgrade-on-risk) as automatic mitigation |
CRP-Safety-Policy |
Mitigation action log |
| Art. 10 | Data governance | CKF fact provenance tracking, source registry, PII detection | CRP-Compliance-GDPR-PII, CRP-Memory-CKF-Community |
Data governance report |
| Art. 11 | Technical documentation | CRP Comply auto-generates Art. 11 documentation from audit trail | CRP-Compliance-Audit-Trail-URI |
Technical Documentation (auto-generated) |
| Art. 12 | Record-keeping | HMAC-chained audit trail with 30+ event types, tamper-evident | CRP-Provenance-HMAC, CRP-Compliance-Audit-Trail-Id |
Complete audit trail export |
| Art. 13 | Transparency | Attribution analysis per-call, grounding percentage disclosed | CRP-Safety-Attribution, CRP-Safety-Grounding-Pct |
Transparency report |
| Art. 14(1) | Human oversight capability | Configurable oversight modes (auto/human-review/halt/log-only) | CRP-Safety-Oversight-Mode |
Oversight event register |
| Art. 14(3)(a) | Ability to fully understand AI system | CRP Visualise provides live session tracing and provenance DAG | — (product feature) | Session replay reports |
| Art. 14(4)(a) | Ability to not use AI system | halt-on CRITICAL immediately stops response delivery |
CRP-Safety-Retry-After: oversight-required |
Halt event log |
| Art. 17 | Quality management system | Quality Tier (S–D) tracking, quality degradation detection, RQA pipeline | CRP-Context-Quality-Tier, CRP-Quality-Score |
Quality management evidence |
| Art. 52 | Transparency obligations (limited risk) | CRP-Compliance-EU-AI-Act: LIMITED classification |
CRP-Compliance-EU-AI-Act |
Transparency obligation checklist |
| Art. 64 | Logging obligations | 30+ event types in HMAC-chained audit trail | CRP-Provenance-HMAC, CRP-Provenance-Chain-Integrity |
EU AI Act logging compliance report |
| Art. 73 | Incident reporting | Safety violations (CRITICAL risk) auto-reported via report-uri webhook |
CRP-Safety-Report-URI |
Incident report template |
Coverage: 16/16 mapped articles. CRP provides technical evidence for all applicable EU AI Act requirements for high-risk AI systems.
2. GDPR (Regulation 2016/679)¶
| Article | Requirement | CRP Feature | CRP Header |
|---|---|---|---|
| Art. 5(1)(a) | Lawfulness, fairness, transparency | Attribution analysis, provenance chain | CRP-Safety-Attribution |
| Art. 5(1)(c) | Data minimisation | CRP-Context-Cache: no-store prevents CKF persistence of PII |
CRP-Context-Cache |
| Art. 5(1)(d) | Accuracy | DPE fidelity verification, fabrication detection | CRP-Provenance-Fidelity-Score, CRP-Safety-Fabrications |
| Art. 5(1)(f) | Integrity and confidentiality | HMAC chain integrity, TLS 1.3, encryption at rest | CRP-Provenance-Chain-Integrity |
| Art. 13/14 | Information to data subject | Attribution type disclosed per-call | CRP-Safety-Attribution |
| Art. 17 | Right to erasure | CKF fact deletion, HMAC chain truncation | — (operational) |
| Art. 22 | Automated decision-making | Oversight mode enforcement, reproducibility seed | CRP-Safety-Oversight-Mode, CRP-LLM-Reproducibility-Seed |
| Art. 25 | Data protection by design | CRP architecture embeds privacy controls (no-store, PII detection, data residency) | CRP-Compliance-Data-Residency |
| Art. 32 | Security of processing | HMAC-SHA256, HKDF key derivation, AES-256-GCM at rest, TLS 1.3 | — (infrastructure) |
| Art. 35 | DPIA | CRP Comply generates DPIA from protocol data | CRP-Compliance-Audit-Trail-URI |
| Art. 44 | Transfer safeguards | Data residency header enforcement | CRP-Compliance-Data-Residency |
3. ISO/IEC 42001:2023 (AI Management Systems)¶
| Control | Requirement | CRP Feature | CRP Header |
|---|---|---|---|
| A.5.2 | AI policy | Safety Policy directive as code-level policy | CRP-Safety-Policy |
| A.5.4 | Roles and responsibilities | Oversight mode assignment per session/system | CRP-Safety-Oversight-Mode |
| A.6.1.2 | AI impact assessment | CRP Comply auto-generates impact assessment from live data | CRP-Compliance-Audit-Trail-URI |
| A.6.2.2 | AI risk assessment | DPE composite risk scoring per-call | CRP-Safety-Hallucination-Risk |
| A.7.3 | Competence and training | — (organisational, not protocol-level) | — |
| A.8.2 | AI system development | Quality tier tracking, conformance levels | CRP-Context-Quality-Tier |
| A.8.4 | AI system testing | Conformance test suite (CRP-SPEC-014) | — |
| A.9.2 | Monitoring and measurement | Continuous DPE analysis on every call | All CRP-Safety-* headers |
| A.9.3 | Internal audit | HMAC-chained audit trail, CRP Visualise | CRP-Provenance-HMAC |
| A.9.4 | Corrective action | upgrade-on-risk automatic remediation, re-dispatch protocol |
CRP-Safety-Policy |
| A.10.2 | Continual improvement | Quality evolution tracking across windows | CRP-Quality-Score |
| B.2.1 | AI system impact | CRP Comply generates AI system impact reports | CRP-Compliance-ISO-42001 |
Coverage: 11/12 mapped controls (A.7.3 is organisational, not protocol-level).
4. NIST AI RMF 1.0¶
| Function | Category | CRP Feature | CRP Header |
|---|---|---|---|
| GOVERN 1.1 | Legal compliance awareness | EU AI Act, GDPR classification per-call | CRP-Compliance-EU-AI-Act, CRP-Compliance-GDPR-PII |
| GOVERN 1.2 | Accountability structures | Audit trail with per-call provenance, oversight modes | CRP-Safety-Oversight-Mode, CRP-Provenance-HMAC |
| GOVERN 1.5 | Risk management integration | DPE integrated into every AI call | CRP-Safety-Hallucination-Risk |
| MAP 1.1 | Intended purpose documented | AI system registration in Gateway configuration | CRP-Compliance-EU-AI-Act |
| MAP 1.6 | Risk tolerance defined | Safety Policy thresholds, safety budget, CRP-Accept-Risk |
CRP-Accept-Risk, CRP-Agent-Safety-Budget |
| MAP 3.5 | Bias and fairness | — (CRP does not currently assess bias; out of scope) | — |
| MEASURE 1.1 | Appropriate methods selected | DPE 13-stage pipeline as measurement methodology | All CRP-Safety-* |
| MEASURE 2.3 | AI system performance | Quality tier, completeness, flow scoring | CRP-Context-Quality-Tier, CRP-Quality-Score |
| MEASURE 2.5 | Trustworthiness evaluation | Composite risk scoring with regulatory amplifiers | CRP-Safety-Hallucination-Score |
| MEASURE 2.6 | Validation against requirements | Conformance test suite (CRP-SPEC-014) | — |
| MANAGE 1.1 | Risk prioritisation | Risk classification (CRITICAL/HIGH/MEDIUM/LOW) | CRP-Safety-Hallucination-Risk |
| MANAGE 2.2 | Monitoring frequency | Continuous (every AI call) | All CRP-Safety-* |
| MANAGE 3.2 | Incident response | CRITICAL halt + webhook reporting + audit trail | CRP-Safety-Report-URI |
| MANAGE 4.1 | Decommissioning | CKF erasure, session termination, data export | — (operational) |
Coverage: 13/14 mapped categories. MAP 3.5 (bias/fairness) is out of CRP's current scope.
5. SOC 2 Type II Trust Service Criteria¶
| Criterion | Requirement | CRP Feature |
|---|---|---|
| CC6.1 | Logical access controls | CRP API key authentication, scoped keys, mTLS |
| CC6.3 | Access control enforcement | Session token scope validation, API key binding |
| CC7.1 | Detection of unauthorized activities | HMAC chain integrity verification (BROKEN = tampering) |
| CC7.2 | Monitoring of system operations | Continuous DPE analysis, 30+ event types in audit trail |
| CC7.4 | Incident management | CRITICAL risk halt, webhook reporting, CRP Comply alerts |
| CC8.1 | Change management | CKF state hash (ETag), version tracking, fact lifecycle |
| CC9.1 | Risk mitigation | Safety Policy enforcement, safety budget, oversight modes |
| A1.2 | Availability monitoring | Gateway health checks, session token expiry management |
6. Australian AI Ethics Framework (CSIRO/DISR)¶
| Principle | Requirement | CRP Feature | CRP Header |
|---|---|---|---|
| P1 Human-centred values | Human oversight capability | CRP-Safety-Oversight-Mode |
|
| P2 Fairness | — (bias detection out of current scope) | — | |
| P3 Privacy | PII detection, no-store cache, data residency | CRP-Compliance-GDPR-PII, CRP-Compliance-Data-Residency |
|
| P4 Reliability & safety | DPE risk scoring, safety budget, halt enforcement | CRP-Safety-Hallucination-Risk, CRP-Agent-Safety-Budget |
|
| P5 Transparency | Attribution analysis, provenance chain, CRP Visualise | CRP-Safety-Attribution, CRP-Provenance-Report-URI |
|
| P6 Contestability | HMAC chain enables audit replay, reproducibility seed | CRP-LLM-Reproducibility-Seed |
|
| P7 Accountability | Complete audit trail, per-call evidence, CRP Comply | CRP-Compliance-Audit-Trail-URI |
|
| P8 Human oversight | Configurable oversight modes, safety budget escalation | CRP-Safety-Oversight-Mode |
7. Cross-Regulation Coverage Summary¶
| CRP Feature | EU AI Act | GDPR | ISO 42001 | NIST RMF | SOC 2 | AU Ethics |
|---|---|---|---|---|---|---|
| DPE risk scoring | Art. 9 | Art. 5(1)(d) | A.6.2.2 | MEASURE 2.5 | CC9.1 | P4 |
| HMAC audit chain | Art. 12, 64 | Art. 5(1)(f) | A.9.3 | GOVERN 1.2 | CC7.1 | P7 |
| Human oversight modes | Art. 14 | Art. 22 | A.5.4 | MANAGE 3.2 | — | P1, P8 |
| Safety Policy | Art. 9(4) | — | A.5.2 | MAP 1.6 | CC9.1 | P4 |
| PII detection | — | Art. 5(1)(c) | — | — | — | P3 |
| Data residency | — | Art. 44 | — | — | — | P3 |
| CRP Comply evidence | Art. 11 | Art. 35 | A.6.1.2 | — | CC7.2 | P7 |
| Quality scoring | Art. 17 | — | A.10.2 | MEASURE 2.3 | — | P4 |
| Safety budget | — | — | — | MAP 1.6 | — | P4 |
| Attribution analysis | Art. 13 | Art. 13 | — | MEASURE 1.1 | — | P5 |
Copyright © 2025–2026 AutoCyber AI Pty Ltd. Licensed under CC BY 4.0. CRP™ is a trademark of AutoCyber AI Pty Ltd.