Перейти к содержанию

06. Compliance

RENAR is a requirements-engineering standard; it is not a compliance standard in its own right. But RENAR provides the traceability infrastructure that makes conformance to other standards (ISO 27001, GDPR, FZ-152, EU AI Act, and others) automatically verifiable. This chapter is a mapping of RENAR artifacts onto 8 key compliance frameworks, plus self-assessment checklists, plus a list of auto-generated artifacts for the auditor.

Prerequisites: RENAR Core, reference/02-schemas.md, reference/03-ai-risk-register.md.


1. Compliance principles in RENAR

1.1 Compliance frontmatter. Every requirement with a regulatory rationale MUST carry a compliance field in its frontmatter — you cannot trace conformance through an external table that is not linked to the artifact. The field is repeatable (one requirement MAY close controls across several standards):

compliance:
  - { standard: "ISO 27001:2022", control: "A.5.34", rationale: "Privacy and protection of PII" }
  - { standard: "GDPR", article: "Art.32", rationale: "Security of processing" }
  - { standard: "FZ-152", article: "Art.19", rationale: "Measures to ensure the security of personal data" }

1.2 Data classification. Every BR/SR that operates on data MUST declare a classification. When contains-pii: true, SR-encryption + SR-audit-log + SR-erasure automatically become mandatory.

data-classification:
  contains-pii: true              # GDPR / FZ-152 trigger
  contains-financial: false       # PCI-DSS trigger
  contains-health: false          # HIPAA / FZ-152 special categories
  contains-children-data: false   # COPPA trigger
  retention-days: 1095
  data-residency: ["RU", "EU"]

1.3 Traceability as the audit foundation. The chain BR → SR → SPEC → TC → implementation → CI run is itself the audit trail. The auditor opens the coverage report and sees: which requirements close a control; which TCs verify them; last-run.date; requirement-version. Audit time: 1–2 days instead of 2–3 weeks. The reconciliation hook (drift detection) guarantees that the evidence has not gone stale between audits.


2. ISO/IEC 27001:2022 — Information Security

ISO 27001:2022 Annex A control RENAR artifact
A.5.7 Threat intelligence SPEC-SEC with a threat model; SR with compliance: A.5.7
A.5.8 Information security in project management RENAR itself as process compliance
A.5.34 Privacy and protection of PII SR with encryption + data-classification.contains-pii: true
A.6.3 Information security awareness The onboarding process includes reading RENAR
A.8.1 User endpoint devices SR with device requirements (if in scope)
A.8.5 Secure authentication SR-AUTH-* + SPEC-SEC with an authentication flow
A.8.7 Protection against malware SR + adversarial review (the AI critic)
A.8.10 Information deletion SR with deletion logic + retention-days
A.8.16 Monitoring activities SR with logging + SPEC-OPS audit-log
A.8.24 Use of cryptography SR with an explicit crypto algorithm + ISO 25010 Security
A.8.25 Secure development life cycle SENAR + RENAR as evidence
A.8.28 Secure coding TC with security checks; SPEC-SEC with STRIDE coverage
A.12.1.2 Change management (legacy 27001:2013) Delta-TZ + Impact Analysis (§7.6)

Audit deliverable. Auto-generates a conformance report: scope (BR/SR/TC counts with compliance.iso27001) + Coverage by Annex A (control ↔ mapped SR ↔ status). The normative mechanism is capability V4 reporting from versioned artifacts.


3. GDPR (Regulation EU 2016/679)

GDPR Article RENAR artifact
Art.5 Principles BR explicitly states the lawful basis
Art.6 Lawfulness BR with gdpr.lawful-basis: consent / contract / legal-obligation / vital-interests / public-task / legitimate-interests
Art.7 Conditions for consent SR with a consent-management workflow
Art.13 Information to data subject SPEC-UI with a privacy-notice screen
Art.15 Right of access SR with export-user-data
Art.16 Right to rectification SR with edit-user-profile
Art.17 Right to erasure SR with delete-user-data + propagation to linked entities
Art.18 Right to restriction SR with suspend-processing
Art.20 Right to data portability SR with export in a machine-readable format
Art.25 Data protection by design and by default data-classification is mandatory at the BR level
Art.30 Records of processing activities Auto-generated from BRs with contains-pii
Art.32 Security of processing SR with encryption + access control
Art.33 Notification of personal data breach SR with a breach-notification workflow + a 72h timer
Art.35 DPIA For high-risk — a separate dpia/<feature>.md

GDPR frontmatter:

gdpr:
  data-categories: ["identification: email, name", "contact: phone, address", "behavioral: usage logs"]
  lawful-basis: contract              # Art.6
  retention-period-days: 1095
  cross-border-transfer: false        # true → SCCs or an adequacy decision are mandatory
  dpia-required: false                # true → link to the DPIA
  data-subject-rights: [access, rectification, erasure, portability]   # Art.15/16/17/20

DPIA. For high-risk processing — <system>.req/dpia/DPIA-NN-<slug>.md with frontmatter (id, type: DPIA, gdpr-article: 35, related-br[], risks-identified, mitigations) and sections: the processing operation, necessity, risks, mitigation measures, and sign-off with the DPO.


4. FZ-152 "On Personal Data" (RF)

FZ-152 Article RENAR artifact
Art.5 Processing principles BR with a stated purpose (business-context.business-goal)
Art.6 Processing conditions BR with a lawful-basis
Art.9 Consent to processing SR with consent management
Art.13.1 Storage within the RF data-classification.data-residency: ["RU"] for Russian clients
Art.14 Right of access to information SR with export-user-data
Art.15 Right to rectification SR with edit-user-profile
Art.16 Deletion SR with delete-user-data
Art.18 Operator obligations The audit trail via RENAR traceability
Art.18.1 Localization of RF citizens' personal data data-residency is mandatory
Art.19 Protection measures SR with encryption + access control + ISO 25010 Security
Art.21 Breach notification SR with a breach-notification workflow
Art.22 Notification to Roskomnadzor operational, outside RENAR scope

Data residency enforcement. For projects with Russian clients, data-residency is mandatory. The substrate hook checks: when data-residency: ["RU"], all upstream SRs (storage, backups, replication) MUST have evidence of storage within the RF; adding a foreign jurisdiction for an RU-resident BR → a block at QG-0.


5. EU AI Act (Regulation EU 2024/1689)

The EU AI Act classifies AI systems by risk level. RENAR requires that the class be stated explicitly:

ai-act:
  risk-class: limited                 # prohibited | high | limited | minimal
  rationale: "Generative AI for document drafting; no autonomous decisions affecting users' legal rights"
  high-risk-domain: false             # true → a conformity assessment is required
  general-purpose-ai: true            # GPAI — Claude / GPT, etc.

High-risk AI requirements (Art.9–15) — for the high class (financial scoring, hiring, public services, medical diagnostics):

AI Act requirement RENAR artifact
Art.9 Risk management system SPEC-AI + AI risk register (reference/03)
Art.10 Data and data governance eval-datasets/ with provenance + spot-check
Art.11 Technical documentation SPEC-AI + ISO/IEC 5338 conformance (§7)
Art.12 Record-keeping tool_event audit + ai-provenance
Art.13 Transparency to users SPEC-UI with an indication of AI processing
Art.14 Human oversight One-click approval + spot-check at QG-0 / QG-2
Art.15 Accuracy, robustness, cybersecurity Eval-tests (tc-type: eval) + adversarial review

GPAI obligations (Art.51–55). If a General-Purpose AI Model (Claude, GPT, Gemini) is used: ai-provenance in the frontmatter of any AI-generated artifact (model id, version, prompt hash); the model's technical document (if it is a GPAI with systemic risk) — an external document + a reference from SPEC-AI.


6. NIST AI RMF 1.0 (US jurisdiction)

NIST AI RMF Function RENAR mechanism
Govern RENAR + SENAR roles + compliance frontmatter
Map BR with business-context, data-classification, ai-act.risk-class
Measure Eval-tests with metric thresholds + RENAR metrics (Hallucination Rate, Coverage Velocity)
Manage Lifecycle with QGs + reconciliation hook + AI risk register

RMF-specific extensions (optional for US projects; not in conflict with ISO/IEC 23894 §7):

nist-ai-rmf:
  applicable: true
  govern: { role: "AI Governance Lead", policy-link: "..." }
  map: { use-case-category: "content-generation", affected-stakeholders: ["clients", "internal-team"] }
  measure: { metrics-tracked: ["accuracy", "fairness", "robustness"], baseline-run-id: "eval-2026-04-01" }
  manage: { risk-tolerance: "low", incident-response-plan: "<link>" }

7. ISO/IEC 23894:2023 — AI Risk Management

Guidance on AI risk management. It does not certify, but it provides a structured framework for managing the risks of AI systems. Compatible with NIST AI RMF and the EU AI Act.

ISO/IEC 23894 clause RENAR artifact
§6.4.1 Risk identification AI risk register (reference/03)
§6.4.2 Risk analysis SPEC-AI with a risk-assessment section
§6.4.3 Risk evaluation A threshold for each risk in eval-tests
§6.4.4 Risk treatment SR-mitigations + post-mitigation evaluation
§6.4.5 Communication and consultation RACI matrix (05-safe-comparison §9)
§6.4.6 Monitoring and review Reconciliation hook (drift detection)

Conformance criteria: every AI-generated artifact carries ai-provenance; for each AI use case, SPEC-AI documents the identified risks (hallucination, bias, robustness), the mitigations (eval thresholds, adversarial review, human-in-the-loop), and the residual risks; the risk register is updated whenever SPEC-AI changes (reconciliation catches the drift).


8. ISO/IEC 5338:2023 — AI System Life Cycle Processes

An extension to ISO/IEC/IEEE 12207. A convenient framework for AI Act Art.11 technical documentation.

ISO/IEC 5338 process RENAR stage
Stakeholder needs and requirements definition TZ + ADAPT + BR
Architecture definition SPEC-ARCH + SPEC-AI
Design definition SPEC-AI (model card, prompt design, RAG)
Implementation TR + implementation
Verification TC (tc-type: eval for AI)
Validation Acceptance TC + spot-check
Operation Reconciliation hook (drift detection, §4.11)
Maintenance Delta-TZ + ADAPT iteration
Disposal Retirement workflow (outside Core RENAR scope)

Conformance evidence: SPEC-AI for each AI use case with a full model card; eval-tests bound to SPEC-AI via verifies[]; ai-provenance recorded; drift detection active (V6).


9. PCI-DSS v4.0 — Payment Card Industry Data Security Standard

If the project handles payment-card data (CHD):

PCI-DSS v4 requirement RENAR artifact
Req.1 Network security controls SPEC-OPS with network segmentation
Req.3 Protect stored CHD contains-financial: true + SR with encryption-at-rest
Req.4 Encrypt CHD transmission SR with TLS + SPEC-API requirements
Req.6 Develop secure systems RENAR + SENAR as process compliance
Req.7 Restrict access by need to know SR with RBAC + SPEC-SEC access-control matrix
Req.8 Authenticate access SR-AUTH-* + MFA where mandatory
Req.10 Log and monitor access SR with an audit trail + SPEC-OPS observability
Req.11 Test security regularly TC tc-type: security + pen-test (outside RENAR scope)
Req.12 Information security policy RENAR + SENAR as a documented policy

CHD scope minimization. The substrate hook checks that new SRs with contains-financial: true explicitly justify the need to store CHD — without that justification, the requirement is halted at QG-0.


10. Self-assessment checklists

Short yes/no checklists for compliance teams. The full printable kit for a RENAR manifest — reference/08.

ISO 27001:2022 — every BR with contains-pii: true has an SR closing A.5.34; the SoA is documented; every in-scope control has ≥1 verifying SR; the coverage report is green; an auditor goes from a control to a passing TC in under 5 minutes.

GDPR — all PII is in the Records of Processing (auto-generated); the lawful basis is stated per processing activity; data-subject rights Art.15–22 are verifiable through SR + TC; a DPIA is performed for high-risk; cross-border transfers are justified (SCCs/adequacy).

FZ-152 — requirements with PII of RF citizens carry data-residency: ["RU"]; the substrate hook checks the upstream storage SRs; consent is implemented in an SR with TC evidence; the Art.19 protection measures are covered by an SR with a passing TC.

EU AI Act — every SPEC-AI carries ai-act.risk-class; for high: Art.9–15 evidence is in the substrate; human oversight at QG-0/QG-2 + spot-check; technical documentation is auto-generated; ai-provenance for GPAI components.

NIST AI RMF — all 4 functions (Govern/Map/Measure/Manage) have evidence; an AI Governance Lead is defined in roles; the eval baseline is recorded; an incident-response plan is linked to SPEC-AI.

ISO/IEC 23894 — the AI risk register is filled out and linked to SPEC-AI; every risk has a mitigation in an SR; residual risks are accepted by the owner; V6 is active.

ISO/IEC 5338 — SPEC-AI for each AI use case with a model card; eval-tests bound via verifies[]; ai-provenance recorded; V6 is active.

PCI-DSS — SRs with contains-financial: true have encryption (at-rest + in-transit); the CHD scope is minimized; RBAC is in SPEC-SEC; the audit-trail SR covers all CHD-touching flows; security TCs run regularly.

10.9 RENAR-conformance self-assessment (in an hour)

A project declares its own RENAR-conformance level via the RENAR-CONFORMANCE.yaml manifest (standard/13 §13.4). Quick checklist (yes/no, single pass):

  • The requirements substrate provides all of V1–V6 (§11.4.1 — Notion/Google Docs do not qualify).
  • Each TZ has exactly one approved ADAPT with dual signature.
  • All 9 SPEC types are supported natively (a declaration of "type not used" is permitted, "not supported" is not).
  • Every normative statement covered by a TC has a paired negative TC.
  • QG-0 / QG-1 / QG-2 are declared required.
  • The manifest contains renar-version + senar-version + level + confirmation of the mandatory clauses (reference/08 §14).
  • assessment-date is recent, next-assessment-due is not overdue.

All 7 are yes → the project is conformant to the declared level. At least one no → the manifest is not issued (§13.5.2).

A filled-in example (RENAR-2, self-assessment) — full schema in §13.4.2:

renar-version: "1.0"
senar-version: "1.0"
manifest-version: 1
manifest-id: "CFM-2026-014"
level: "RENAR-2"
level-target: "RENAR-3"
assessment-mode: "self"
assessment-date: "2026-05-20"
assessor: { id: "architect-team-lead", role: "architect", signature-ref: "<pointer>" }
next-assessment-due: "2026-08-20"
mandatory-clauses-confirmed:
  sot-inversion: true
  substrate-v1-v6: { v1: true, v2: true, v3: true, v4: true, v5: true, v6: true }
  adapt-per-tz: true
  spec-types-closed-list: true
  tc-pos-neg-pairing: true
  quality-gates-closed-list: true
  closed-lists-backward-findings: true
quality-gates: { qg-0: required, qg-1: required, qg-2: required, qg-3: absent, qg-4: absent }
spec-types-supported: ["SPEC-ARCH","SPEC-API","SPEC-DATA","SPEC-INT","SPEC-PROC","SPEC-UI","SPEC-AI","SPEC-SEC","SPEC-OPS"]
exceptions: []
replaced-by: null

11. Auto-generated compliance artifacts

Artifacts generated from existing requirements for auditors:

Artifact Source Contents
Records of Processing (GDPR Art.30) BR/SR with contains-pii: true Data categories, lawful basis, retention, recipients
Statement of Applicability (ISO 27001) BR/SR with compliance: ISO 27001:2022 Annex A controls in/out of scope, justification
Data Inventory All data-classification records Categories, residency, retention, owners
AI System Cards SPEC-AI Model card in NIST AI RMF / AI Act format
Audit-trail report Traceability BR → SR → SPEC → TC → last-run The chain from control to evidence
DPIA Index The dpia/ folder A list of DPIAs + their DPO sign-off status
AI Risk Register Snapshot AI risk register All identified risks + mitigations + residual

Substrate-native reporting: aggregation of evidence from versioned artifacts (V4) into a compliance report on the assessor's request.

Evidence pack — templates (informative; full E2E — guide/09 §E3). GDPR Art.15 trace bundle — a Control | BR/SR | SPEC | TC | last-run table + lawful basis + retention + a link to the manifest. FZ-152 Art.14 — a mapping table Requirement ↔ RENAR artifact ↔ Evidence field. ISO 29148 trace excerpt — reference/07 (fill in "RENAR frontmatter" for each in-scope SR). Pre-audit self-assessment — reference/08 §14.


12. What RENAR does NOT cover in compliance

RENAR is infrastructure for compliance, but not a substitute for: the DPO (Data Protection Officer — a human role with legal accountability); legal review of contracts and privacy notices; pen-testing the implementation; bug bounty / vulnerability management; physical security (data centers, the office); operational security (key rotation, incident response, the monitoring runbook); cyber insurance; regulatory filings (Roskomnadzor notifications, GDPR registration with a DPA, AI Act conformity assessment).

RENAR helps you do compliance during requirements engineering. Operational compliance consists of separate processes that reference RENAR artifacts as evidence.


13. Resolved decisions for v1.0

  • Compliance frontmatter — conditionally mandatory. By default it is optional; it is mandatory when contains-pii: true (GDPR/FZ-152 trigger) or contains-phi: true (HIPAA trigger). An artifact with PII and no compliance frontmatter is non-conformant (§13.3 extensions via a manifest declared-stricter).
  • DPIA — mixed model. The formal DPIA is an external document (legal owns it); the RENAR artifact dpia/<slug>.md holds a machine-readable summary + a pointer to the formal document. Separation of concerns while preserving traceability.
  • Data residency — outside RENAR scope. Per §1.3 (3), tech stack / infra are out of scope. Enforcement is via DevOps-level controls (network policies, cloud regions). RENAR records the requirement (data-residency.region: eu-west), not the enforcement.
  • Custom industry frameworks (CB RF fintech, HIPAA healthcare, etc.) — separate documents (industry-specific addenda as guide/06-compliance-<industry>.md or external documents; they MAY be declared-stricter on top of RENAR conformance).

Deferred to v1.1 (phase-8 backlog): auto-export of evidence for Schrems II and EU ↔ RF transfers — partially addressed by the cross-border-transfer flags and links to SCCs, but full automation is unattainable (which SCCs apply is a lawyers' call). Owners: the legal-tech / adapters pairing.


14. Relationship to other chapters