MRIS assesses the 93 controls of ISO/IEC 27001:2022 Annex A (ISO/IEC 27002:2022) against AI-accelerated attacks — and names concrete compensating controls for every gap.
Mythos (Anthropic) was the first frontier AI model whose capabilities in an attacker’s hands fundamentally changed the security landscape. MRIS uses it as the reference threat model: every control is measured against an attacker with Mythos-level capabilities.
The time between a known vulnerability and its exploitation is shrinking toward zero.
Minutes, not days, pass between initial access and damage.
Attacks run in parallel and distributed to overwhelm the response.
Attack capability no longer depends on the actor's resources.
In the work's gap analysis (ch. 8), these four shifts condense into seven systematic gaps in ISO/IEC 27002:2022 – requirements that other frameworks already mandate and that deliver hard protective effect under Mythos conditions:
Thirteen established security works, one question: what does a work contribute to an ISMS under the current Gen-AI threat landscape? The X-axis measures Mythos/Gen-AI coverage, the Y-axis implementation readiness and auditability.
| Work | Category | Gen-AI coverage (X) | Implementation readiness (Y) |
|---|---|---|---|
| ISO/IEC 27002 + MRIS (hardened framework) | Combination | 9.8 | 9.0 |
| MRIS for ISO 27001 (v1.8)* | Hardening layer | 9.8 | 7.9 |
| Mythos-Ready (CSA/SANS/OWASP) | Community work | 8.8 | 4.6 |
| OWASP Agentic Top 10 (ASI) | Community work | 7.8 | 3.8 |
| MITRE ATT&CK v19 / ATLAS | Knowledge base | 6.8 | 6.4 |
| BSI C5:2026 | Norm / standard | 5.4 | 9.0 |
| NIST AI RMF + GenAI Profile | Practice framework | 5.0 | 5.8 |
| ISO/IEC 42001:2023 | Norm / standard | 4.6 | 6.2 |
| CISA Zero Trust MM v2.0 | Practice framework | 4.0 | 6.8 |
| NIST CSF 2.0 | Practice framework | 3.6 | 7.0 |
| CSA Cloud Controls Matrix v4 | Norm / standard | 3.5 | 8.6 |
| CIS Controls v8.1 | Practice framework | 3.2 | 7.6 |
| BSI IT-Grundschutz | Norm / standard | 3.0 | 8.4 |
| ISO/IEC 27002:2022 | Norm / standard | 2.2 | 9.0 |
* MRIS single position per the work’s figure; Table 1 of the work lists MRIS within the combination.
Reading on three levels: (1) ISO/IEC 27002 alone is maximally certifiable, yet as a 2022 work it has no answer to the Gen-AI threat landscape. (2) MRIS alone reaches the highest Gen-AI coverage in the field (9.8) but is deliberately not a standalone ISMS. (3) The combination – the golden star at 9.8 / 9.0 – is the decision-relevant unit in practice: highest Gen-AI coverage and highest execution strength. Certification still targets the ISMS under ISO/IEC 27001, with an SoA extended by the 13 MHC.
Qualitative expert scoring (0–10 scale, uncertainty band ±0.3 on X, ±0.4 on Y, shown as error bars on the MRIS point). Methodology, scoring rationale and rubric are fully disclosed in the work. Assessment status: July 2026.
Each of the 93 controls from ISO/IEC 27001:2022 Annex A is assessed against the AI threat landscape. The result feeds directly into your risk assessment under ISO/IEC 27005.
Select a control from ISO/IEC 27002:2022. MRIS shows the flanking Mythos-Hardening Controls (MHC) that close the gap.
Mappings per MRIS, Annex A/C. Flanked controls per v1.8 – including proactively flanked robust controls.
Number of controls each MHC flanks. Prioritization starts with the broadest impact.
New in v1.8 · July 2026 Mappings extended: A.5.22 (monitoring of supplier services) and A.8.15 (logging) added to the MHC mapping · proactive flanking of robust controls introduced (A.5.9 → MHC-13) · market comparison with four-quadrant positioning added.
Complete assessment of the 93 controls from ISO/IEC 27001:2022 Annex A, with gap analysis and MHC catalog.
DownloadPrioritization with Threat Priority Score, roadmap and RACI for the 13 MHC.
DownloadMRIS and the Implementation Guide are working aids for assessing the effectiveness of existing security controls. They assume an established ISMS and do not replace it. They do not constitute legal, compliance or certification advice, make no claim to completeness and establish no warranty. Use is at your own responsibility. Named standards, frameworks and trademarks belong to their respective owners.