Untitled
A structured library of deep-dives covering the standards, regulations, policy instruments, and professional frameworks that make up the AI assurance landscape, with an Australian focus and international context.
For an assessment of coverage and the backlog of planned additions, see AI-Assurance-Coverage-Gap-Analysis.md.
Navigate by Theme
By geography
| Geography | Key documents |
|---|---|
| Australia — regulators | APRA · APRA AI obligations (Apr 2026) · ASIC · OAIC · TGA · ACMA · ACCC |
| Australia — policy | AI Ethics Principles, Voluntary AI Safety Standard, National AI Plan, AI Safety Institute |
| Australia — market | AI assurance providers & certification bodies · GRC platforms · APRA-regulated entity landscape |
| Australia — consulting firms | Deloitte AU strategy · KPMG AU (ISO 42001 first, Oct 2024) · EY AU (AI Sentiment Index) · PwC AU (Assurance for AI launch) |
| EU | EU AI Act (extraterritorial obligations for AU firms) · EU AI Act deep-dive |
| US | US AI regulations (FTC, CFPB, California) · McKinsey (SR 11-7, financial services) · PwC US (AICPA Assurance for AI) |
| UK & Singapore | UK & Singapore AI frameworks |
| Other markets | Canada, NZ, Hong Kong, Japan, UAE |
| Global (standards) | ISO/IEC SC 42 family · NIST AI RMF · OWASP AISVS + TEA Platform |
| Global (consulting) | Cross-firm synthesis · BCG global RAI surveys · Deloitte global publications |
By industry
| Industry | Key documents |
|---|---|
| Financial services (APRA-regulated) | APRA obligations (CPS 220/230/234) · APRA-regulated entities landscape · Deloitte AU strategy · McKinsey SR 11-7 / model risk · ASIC |
| Healthcare / medical devices | TGA (AI as Software as a Medical Device) |
| Insurance | APRA (prudential: life insurance, general insurance, private health) |
| Government / public sector | EY AU APS guidance · Australian AI Policy Instruments |
| Communications / media | ACMA + eSafety Commissioner |
| Multi-sector | BCG/MIT SMR RAI surveys · PwC Responsible AI Toolkit · KPMG Trusted AI Framework |
By function
| Function | Key documents |
|---|---|
| Regulatory obligations | Australian regulator analyses: APRA · ASIC · TGA · OAIC · ACCC · ACMA; extraterritorial: EU · US · UK/SG |
| External assurance practitioners | ASAE/ISAE 3000 practitioner reference · AUASB/IAASB standards deep-dive · APESB/APES ethics & independence · SOC 2 · PwC Assurance for AI · KPMG ISO 42001 certification |
| Internal audit | IIA Global Internal Audit Standards + AI Auditing Framework · ISACA AAIA certification · ISACA audit standards (ITAF, COBIT) |
| Advisory / strategy | Consulting firms directory → · Cross-firm synthesis · Deloitte AU strategy |
| Professional certifications | ISACA: CISA, CISM, CRISC, CGEIT, CDPSE, AAIA, AAISM · ISACA AAIA deep-dive · Australian professional bodies |
By theme
| Theme | Key documents |
|---|---|
| Trustworthy / responsible AI frameworks | Cross-firm synthesis · Deloitte Trustworthy AI Framework · KPMG Trusted AI (10 pillars) · PwC Responsible AI Toolkit · EY Nine-Attribute Framework · BCG RAI Leader Blueprint |
| ISO 42001 certification chain | ISO/IEC 42001 deep-dive · KPMG (world’s first, AU Oct 2024) · ISO 42006 (certification body requirements) |
| Agentic AI assurance | GenAI + agentic AI cross-cutting synthesis · Deloitte TAI across AI spectrum · McKinsey observability gap · BCG FAST framework · OWASP AISVS agentic controls |
| Model validation | McKinsey SR 11-7 extension + Derisking AI · Deloitte: Adapting Model Validation · GenAI assurance cross-cutting |
| GRC platforms | GRC Platforms Market Scan |
| Offensive vs. defensive governance | Deloitte offensive/defensive analysis · BCG RAI as value creation |
How the Library Is Organised
| Layer | Question | Key Documents |
|---|---|---|
| Technical & definitional standards | What is an AI system and what does technically “good” look like? | ISO/IEC 22989, 23053, 5338, 5259, TR 24027 |
| Management system & risk standards | How should organisations govern AI (and get certified)? | ISO/IEC 42001, 42005, 42006, 23894, 27001, 27701 |
| Regulatory obligations | What must organisations comply with? | EU AI Act; Australian regulator analyses; extraterritorial analyses |
| Policy instruments | What do governments expect even where not mandated? | Australian AI Policy Instruments |
| Professional & engagement standards | Who performs assurance and under what obligations? | APESB/APES, AUASB/IAASB (ASAE/ISAE), IIA, ISACA, SOC 2, Australian professional bodies |
| Industry initiatives | What industry-led frameworks, benchmarks, and consortia can be leveraged? | Industry Bodies deep-dive; Frontier AI Industry Initiatives |
| Market landscape | Who provides assurance and certification in Australia? What software infrastructure supports AI governance and assurance? | AI Assurance Market Scan; GRC Platforms Market Scan |
| Consulting firm AI assurance research | What are the major consulting and professional services firms offering, how do their frameworks compare, and what are the strategic implications for Deloitte in Australia? | KPMG, PwC, EY, BCG, McKinsey, Deloitte research compendiums and critical syntheses; cross-firm comparative synthesis; Deloitte Australia strategy |
Contents
June 2026 restructure: the knowledge base was split into three sibling directories. This directory (
AI-assurance/) holds professional assurance standards, regulator obligations, and the consulting-firm market layer. The ISO/IEC SC 42 family, NIST AI RMF, the EU AI Act, and industry-body standards now live in../AI-Governance/standards/; MITRE ATLAS, the NCSC/CISA/ASD guidelines, and ISO/IEC 27001/27701 now live in../AI-security/. Links below point to those locations where a document has moved.
Overviews and Cross-Cutting
- AI-Standards-and-Assurance-Overview.md — meta-overview of standards bodies, assurance approaches, and topic-specific assurance techniques
- ISO-SC42-AI-Standards-Family.md — the ISO/IEC JTC 1/SC 42 catalogue, how the standards fit together, and deep-dive availability (in
../AI-Governance/standards/) - Industry-Bodies-AI-Standards-Deep-Dive.md — ISACA, IEEE, IAPP, Partnership on AI, BSI, UNESCO, OECD (in
../AI-Governance/standards/) - Frontier-AI-Industry-Initiatives-Deep-Dive.md — Frontier Model Forum, Seoul Frontier AI Safety Commitments, MLCommons AILuminate, CSA AI Controls Matrix and STAR for AI, AI Alliance, ForHumanity, OWASP (in
../AI-Governance/standards/) - NIST-AI-RMF-Deep-Dive.md — NIST AI RMF (AI 100-1), AI 600-1 (GenAI risk profile), and AI 100-2 (adversarial ML taxonomy) as a family document (in
../AI-Governance/standards/) - MITRE-ATLAS-Deep-Dive.md — MITRE ATLAS v5.4 adversarial AI threat knowledge base (tactics, techniques, mitigations, and real-world case studies) (in
../AI-security/) - NCSC-CISA-ASD-Secure-AI-Development-Deep-Dive.md — joint government guidelines for secure AI system development co-authored by 18 national agencies including ASD/ACSC (November 2023) (in
../AI-security/) - OWASP-AISVS-and-TEA-AssurancePlatform-Deep-Dive.md — OWASP AI Security Verification Standard (14-chapter requirement catalogue, agentic and MCP controls) and the Alan Turing Institute TEA Platform (open-source assurance case tooling)
- GenAI-Agentic-AI-Assurance-Cross-Cutting.md — cross-cutting synthesis of GenAI and agentic AI assurance: why traditional validation is insufficient, the current evidence base, candidate criteria, and what a practitioner can and cannot conclude (draws on all five AI directories)
- AI-Assurance-Coverage-Gap-Analysis.md — coverage map, gap register, and backlog
TODO.md— next steps for this knowledge base
ISO/IEC Standards Deep-Dives (now in ../AI-Governance/standards/ and ../AI-security/)
- ISO-IEC-42001-Deep-Dive.md — AI management systems (certifiable)
- ISO-IEC-42005-Deep-Dive.md — AI system impact assessment
- ISO-IEC-42006-Deep-Dive.md — requirements for bodies auditing and certifying AIMS
- ISO-IEC-23894-Deep-Dive.md — AI risk management
- ISO-IEC-23053-Deep-Dive.md — framework for AI systems using machine learning
- ISO-IEC-5338-Deep-Dive.md — AI system lifecycle processes
- ISO-IEC-22989-Deep-Dive.md — AI concepts and terminology
- ISO-IEC-24027-Deep-Dive.md — bias in AI systems and AI-aided decision making
- ISO-IEC-5259-Deep-Dive.md — data quality for analytics and ML (series)
- ISO-IEC-27001-Deep-Dive.md — information security management (in
../AI-security/) - ISO-IEC-27701-Deep-Dive.md — privacy information management (in
../AI-security/)
Regulation and Attestation Frameworks
- EU-AI-Act-Deep-Dive.md — EU Regulation 2024/1689 (in
../AI-Governance/standards/) - SOC2-Trust-Services-Criteria.md — AICPA SOC 2 attestation
Professional Assurance Standards (Professional-Assurance-Standards/)
- README — directory index
- APESB-APES-Standards-Deep-Dive.md — APES 110 Code of Ethics (incl. 2025 technology revisions), APES 320/325, and related standards for AI services
- AUASB-IAASB-Assurance-Standards-Deep-Dive.md — ASAE/ISAE 3000 family, ASQM/ISQM, criteria selection, and AI engagement design
- ASAE-ISAE-3000-Practitioner-Reference.md — practitioner decision guide: engagement types, the five-element test, mandatory vs conditional ASAE 3000 paragraphs, report elements, and practitioner-qualification pathways for AI engagements
- IIA-AI-Auditing-Framework.md — Global Internal Audit Standards, Three Lines Model, and the IIA AI Auditing Framework
- Australian-Professional-Bodies-AI-Guidance.md — AICD/HTI director resources, Governance Institute, Actuaries Institute, ACS, Engineers Australia, and the legal profession (law societies and court protocols)
ISACA (ISACA/)
- README — directory index
- ISACA-Overview.md, ISACA-Audit-Standards.md, ISACA-Certifications.md, ISACA-AAIA-Certification.md
Australian Market Scan and Policy (AI-Assurance-market-scan/)
- AI-Assurance-Market-Scan.md — Australian AI assurance landscape, providers, and certification bodies
- GRC-Platforms-Market-Scan.md — proprietary and open-source GRC software infrastructure for AI governance and assurance: AI-native platforms (Credo AI, Holistic AI, KomplyAI), compliance automation (Vanta, Drata, OneTrust), GRC incumbents (ServiceNow, IBM, Optro), technical assurance vendors, open-source platforms, analyst coverage, and 2024–2026 consolidation
- Australian-AI-Policy-Instruments.md — AI Ethics Principles, Voluntary AI Safety Standard, NAIC Guidance for AI Adoption (AI6), mandatory guardrails proposal, National AI Plan, AI Safety Institute, regulator-specific AI guidance, government assurance frameworks
- Regulator analyses: ASIC, APRA, ACMA, ACCC, OAIC, TGA
- APRA-AI-Assurance-Obligations.md — analysis of APRA’s 30 April 2026 Letter to Industry on AI: AI-specific expectations against CPS 220/230/234, criteria-readiness matrix, and engagement design by obligation
- APRA-Regulated-Entities-Landscape.md — sector rankings of the largest APRA-regulated entities (ADIs, super, insurance) sizing the AI-assurance opportunity
- Extraterritorial analyses: EU, US, UK & Singapore, other markets
Consulting Firm AI Assurance Research (consulting-firms/)
Research compendiums (hyperlinked catalogues of public publications) and critical syntheses for each major firm, plus cross-cutting analysis. See consulting-firms/README.md for thematic navigation by geography, function, and industry.
Research compendiums:
- Deloitte-AI-Assurance-Research.md — 50+ publications; Trustworthy AI™ Framework; AI Institute; algorithm assurance; regional offices (US, UK, AU, Japan, APAC)
- KPMG-AI-Assurance-Research.md — Trusted AI 10-pillar framework; ISO 42001 certification chain (world’s first: AU Oct 2024 → International Dec 2025); University of Melbourne trust study (48,000+, 47 countries); System Cards
- PwC-AI-Assurance-Research.md — 76 publications; Responsible AI Toolkit (6 dimensions); Assurance for AI (June 2025, AICPA standards); AI Trust Dividend (1.7×); Model Edge; Regulatory Pathfinder
- EY-AI-Assurance-Research.md — 64 publications; Trusted AI Framework (9 attributes, 3 governance domains); Responsible AI Pulse Survey Phase 2 (99% of orgs lost money, avg US$4.4M); AI Sentiment Index (AU equal lowest globally, score 52); APS government guidance
- McKinsey-AI-Assurance-Research.md — 48+ publications; AI Trust Maturity Model (5 dimensions, 0–4 scale, longitudinal); SR 11-7 extension for ML/AI; State of AI survey series; The AI Reckoning (Dec 2025)
- BCG-AI-Assurance-Research.md — 53 publications; RAI Leader Blueprint (5 pillars incl. Culture); BCG/MIT SMR RAI survey series; ARTKIT open-source red-teaming; FACET explainability library; FAST framework for agentic AI
Critical syntheses:
- Deloitte-AI-Assurance-Synthesis.md — house view, brand architecture, TAI framework instability, regulatory posture divergence, October 2025 incident, cybersecurity–assurance silo
- Deloitte-TAI-Framework-Across-AI-Spectrum.md — TAI framework applied across predictive ML, GenAI, and agentic AI
- Deloitte-Offensive-vs-Defensive-AI-Governance.md — Japan’s offensive governance concept; governance-as-enabler analysis; quality management analogy; HRO theory
- KPMG-AI-Assurance-Synthesis.md — ISO 42001 first-mover vs. BSI independence; pillar instability; Melbourne survey not integrated into framework
- PwC-AI-Assurance-Synthesis.md — AICPA attestation as structural innovation; Trust Dividend methodology; security gap in Toolkit; self-review threat
- EY-AI-Assurance-Synthesis.md — US$4.4M loss data as most important financial materiality finding; AI Assessments as public policy; governance watchtower; Australia trust crisis
- McKinsey-AI-Assurance-Synthesis.md — AI Trust Maturity Model rigour; SR 11-7 extension; LLM validation gap; QuantumBlack RAI Principles as unverified self-assertion
- BCG-AI-Assurance-Synthesis.md — offensive governance framing; third-party AI risk paradox (53%/55%); ARTKIT/FACET as only open-source governance tools; FAST framework assessment
Cross-cutting analysis:
- AI-Assurance-Firms-Comparative-Synthesis.md — cross-firm comparison of all six: framework architectures, what “assurance” means, active disagreements, shared gaps, Australian market specifics, competitive positioning map
- Deloitte-Australia-AI-Assurance-Strategy.md — strategic positioning for Deloitte in Australia: strengths to emphasise, weaknesses to mitigate, priority moves (ASAE 3000, APS implementation, APRA agentic framework, independence policy), priority action sequence
- Deloitte-Australia-AI-Assurance-Priority-Plan.md — validation plan converting the (public-information-based) strategy into prioritised hypotheses to confirm with Deloitte leadership and test with priority clients, with decision gates per strategic move
Conventions
- Deep-dives:
{Standard-Code}-Deep-Dive.md; regulator analyses:Regulator-Analysis-{Agency}.md; each subdirectory carries aREADME.mdindex - Document structure: header metadata → executive summary → structure/key requirements → implementation guidance → relationships to other standards → checklists → gaps/future developments → sources
- Each document carries a status note; verify fast-moving items (Australian policy, draft ISO standards, IAASB/IIA workstreams) against primary sources before relying on them
Related Directories
../AI-Governance/— internal governance frameworks, controls templates, and risk management artefacts that implement what these standards require../research/tools/— Responsible AI tooling options analyses../ai-governance-tooling/— AI governance tooling ideas and PRDs