Integrated Security and AI Governance Framework
How to run ISO 27001, ISO 42001, SOC 2 Type II, ASD Essentials, the ISM and APRA CPS 234 and 230 as one programme, with one owner, one risk register and one control set, and a 20-step checklist to work through.
Why one framework, not five
Few organisations need just one framework. A typical Australian business answers to several at once: ISO 27001 because customers ask for it, ISO 42001 because AI is now in the product or the workplace, ASD Essentials or the ISM for government work, and APRA's CPS 234 and 230 if it is regulated. Run each as its own project and you get competing assessments, duplicate policies and the same evidence collected four times.
An integrated framework is the governance layer above them: one accountable owner, one risk appetite, one risk register, one control set and one incident process, with each framework read as a mapping onto that control set rather than a separate programme. Most controls satisfy several frameworks at once; the crosswalk below shows where.
How to use this guide
Start with which combination you need, check the crosswalk to see how much overlaps, then work through the 20-step checklist and the example project plan. The order matters more than the pace: drivers and appetite first, certification decisions last, once you know where your gaps are.
Tick items as you complete them. Progress is stored in your own browser and never leaves it.
Physical and personnel security
An integrated programme can also cover physical and personnel security, and for critical infrastructure it must: the SOCI Act's risk management program is all hazards, covering cyber, personnel, supply chain and physical risks. ISO 27001's Annex A and the ISM both include physical and personnel controls, so either spine carries them. Fold them into the same register and incident process rather than running them separately.
Which combination do you need?
Your frameworks follow from who is asking for assurance. Common situations:
SaaS selling to enterprise
ISO 27001 · SOC 2 for US buyers · ISO 42001 if AI is in the product
ISO 27001 is the usual entry ticket for Australian and European buyers. Add ISO 42001 when customers start asking how your AI is governed, as an extension of the same management system.
Selling to Australian government
ASD Essentials / Essential Eight · ISM · IRAP for PROTECTED
Contracts often name an Essential Eight maturity level, which still applies until the buyer updates it. The ISM is the control catalogue, and an IRAP assessment is needed for PROTECTED systems.
APRA-regulated entity using AI
CPS 234 · CPS 230 · ISO 42001 · ISO 27001 as structure
CPS 234 and 230 are obligations; ISO 27001 is a common way to organise the programme that meets them. Treat AI providers as service providers under CPS 230 and AI systems as information assets under CPS 234.
Critical infrastructure
SOCI risk management program · ISO 27001 or ISM · ASD Essentials for OT
The SOCI program is all hazards and expects a recognised cyber framework behind it. ASD Essentials has a separate chapter for operational technology.
Deploying AI agents internally
ASD Essentials · ISM AI controls · ISO 42001
Agents need their own identities, least privilege, human approval for high-impact actions and logging. Our ASD Essentials and AI guide maps these to the ISM controls; ISO 42001 supplies the governance around them.
Startup preparing for enterprise deals
Phase 0 and 1 now · ISO 27001 when a deal requires it
Do the foundation work early (sponsor, appetite, inventory, policies), because it is cheap now and expensive to retrofit. Certify when a specific customer or tender makes it worth the cost.
The crosswalk: one control area, many frameworks
Each row is a control area you build once. The columns show where each framework asks for it. References are to the main clauses and controls, not every one that applies.
| Control area | ISO 27001:2022 | ISO 42001 | ASD Essentials and ISM | SOC 2 (Trust Services Criteria) | APRA CPS 234 and 230 | NIST CSF 2.0 |
|---|---|---|---|---|---|---|
| Governance and accountability | Clause 5 Leadership | Clause 5 Leadership; AI policy (A.2) | Governance and culture; GOV-08 | CC1 Control environment; CC2 Communication | Board accountable for information security (234) and operational risk (230) | Govern (GV) |
| Risk assessment and treatment | Clauses 6.1, 8.2, 8.3 | Clause 6.1; AI system impact assessment (A.5) | Risk-based priority phases | CC3 Risk assessment; CC9.1 | Controls commensurate with threats (234); operational risk management (230) | GV.RM, ID.RA |
| Asset inventory | A.5.9 Inventory of assets | Resources for AI systems (A.4) | Asset identification; ISM-2134 ISM-2135 | CC6.1 (information asset inventory) | Classify information assets by criticality (234); critical operations (230) | ID.AM |
| Identity and access | A.5.15 to A.5.18; A.8.2 | Covered through the ISMS | Identity, credential and access management; least privilege; ISM-2133 ISM-2156 ISM-2157 | CC6.1 to CC6.3 | Information security controls (234) | PR.AA |
| Suppliers and third parties | A.5.19 to A.5.23 | Third-party and customer relationships (A.10) | Supplier assurance; cyber supply chain; ISM-2103 | CC9.2 Vendor and business partner risk | Third-party controls (234); material service providers (230) | GV.SC |
| AI system lifecycle and data | Not specific to AI | AI system life cycle (A.6); data for AI systems (A.7) | ISM-2084 ISM-2086 ISM-2087 ISM-2113 ISM-2158 | CC8.1 Change management, applied to AI systems | Through information asset and service provider controls | See NIST AI RMF |
| Logging and monitoring | A.8.15 Logging; A.8.16 Monitoring | Event logs within the AI life cycle (A.6) | Centralised event logging; ISM-2159 ISM-2114 | CC7.2 System monitoring | Detect incidents (234) | DE.CM |
| Incident management | A.5.24 to A.5.28 | Through the ISMS | Visibility and response | CC7.3 to CC7.5 | Notify APRA of material information security incidents within 72 hours (234) | RS |
| Continuity and resilience | A.5.29, A.5.30 | Through the ISMS | Regular and proven backups, including AI models and data | A1.2, A1.3 Availability (if in scope) | Business continuity and tolerance levels (230) | RC |
| Assurance and improvement | Clauses 9, 10 | Clauses 9, 10 | Effectiveness ratings per action and domain | CC4 Monitoring activities | Test control effectiveness; internal audit (234) | ID.IM |
This crosswalk is CISO AI's own analysis for planning purposes, not an official mapping. Check each reference against the standards before relying on it.
SOC 2 Type II in an integrated programme
SOC 2 is an attestation report under the AICPA's Trust Services Criteria, issued by an independent CPA firm. It is not a certificate. The criteria cover security (always in scope) and, by choice, availability, processing integrity, confidentiality and privacy. A Type I report says your controls were suitably designed at a point in time; a Type II report says they operated effectively over an observation period, typically three to twelve months, which is what most US enterprise buyers ask for.
It fits an integrated programme well, for two reasons:
- Most of the controls are already there. The Trust Services Criteria overlap heavily with ISO 27001's Annex A, so the same control set, mapped in the crosswalk above, serves both. The extra work is mainly in how evidence is collected.
- Evidence must be continuous. Type II tests whether controls ran throughout the period, so evidence has to be collected as you go: access reviews done on schedule, changes approved, incidents handled. A programme that already collects evidence once and tags it to every framework (Phase 3 of the checklist) turns this from a scramble into a report.
What auditors test for AI in a Type II
SOC 2 has no AI-specific criteria. Auditors test the controls you describe in your system description against the same Trust Services Criteria, so if AI is in scope, your AI controls are what gets sampled, across the whole observation window. In practice that looks like this:
| Area | Criteria | Evidence an auditor samples across the period | Related ISM |
|---|---|---|---|
| Access, for people and agents | CC6.1 to CC6.3 | Access reviews done on schedule for training environments, prompt and model registries and vector databases, covering staff and non-human identities, with least privilege enforced and leavers removed. | ISM-2133 ISM-2156 ISM-2093 |
| Model and prompt changes | CC8.1 | Approved change tickets for system prompt and prompt template changes, fine-tuning runs, model version rollouts and retrieval knowledge-base updates, with testing before release. | ISM-2084 ISM-2086 |
| Model providers and subprocessors | CC9.2 | Vendor risk assessments for foundation model providers, and agreements that set out data retention and whether your customer data may be used for training. Most customers now expect a clear no on training. | ISM-2103 |
| Runtime and agent monitoring | CC4, CC7.2 and CC7.3 | Logs and alerts for agent tool calls, rate limit breaches and anomalous outputs that ran throughout the period, and evidence that alerts were followed up. | ISM-2159 ISM-2114 ISM-2090 |
| Retention and deletion | C1 Confidentiality; P Privacy, if in scope | Deletion records showing prompts, outputs, embeddings and fine-tuning datasets were purged as your published commitments say. | ISM-2123 |
The criteria references are the ones these controls most often map to; your auditor will confirm the mapping for your system description. Several of these controls are covered in more depth in our securing AI agents guide.
Two practical points. Start the observation period only once the controls have been running for a few weeks, because exceptions in the window appear in the report. And keep ISO 27001 and SOC 2 audits close together in the calendar where you can, so one evidence set serves both. Australian buyers mostly ask for ISO 27001; add SOC 2 when US customers make it a condition.
Programme owner
CISO or CRO with a cross-framework mandate
Spine
ISO 27001 or NIST CSF 2.0
Mapped to
ISO 42001 · SOC 2 · ASD Essentials · ISM · CPS 234/230
Implementation progress
Phase 0: Foundation and drivers
Formalise the programme with a C-level sponsor and a steering committee drawn from the business, IT, risk, legal and procurement. With several frameworks in play, the sponsor is also who settles which one wins when they pull in different directions.
Write down why you are doing this and who is asking: a customer contract (ISO 27001, SOC 2), government work (ASD Essentials, the ISM, IRAP), a regulator (APRA CPS 234 and 230, SOCI), or AI in your product or operations (ISO 42001). Every framework on your list should trace to a driver, or it comes off the list.
Write measurable appetite statements once, for the whole organisation: tolerance for downtime, record exposure, regulatory findings, and how much autonomy AI systems may have before a person must approve. Separate appetites per framework are how contradictions go unnoticed.
List the AI tools staff use, the AI features switched on in your SaaS platforms, the models you build or fine-tune, and every AI agent with access to your systems. ISO 42001, ASD Essentials and the ISM all start from this inventory, and most organisations find more than they expected.
Phase 1: Design one programme
Pick one structure to organise everything, usually ISO 27001 (certifiable) or NIST CSF 2.0 (outcome-based), and cross-walk every other framework to it. One control set with many mappings beats one control set per framework; the crosswalk below is a starting point.
Set the organisational scope (whole entity, a division, a product) and the technology scope, and decide which AI systems are inside it. Scope is the biggest driver of certification cost, and it is far cheaper to settle now than during an audit.
Systems, data, sites, suppliers, AI models, AI services and agents in one inventory, with their owners. The agent register the ISM asks for (ISM-2134) and ISO 42001's inventory of AI resources are views of this one list, not separate documents.
Assess once, against the spine, and read the results through the crosswalk for every framework on your list. Three separate assessments by three consultants produce three remediation plans that compete for the same budget.
A Tier 1 charter, Tier 2 policies and Tier 3 standards and procedures, including an AI policy (ISO 42001 asks for one) and a general-purpose AI usage policy for staff (ISM-2074). Write policies that describe what you actually do; an aspirational policy is an audit finding waiting to happen.
Phase 2: Build and implement
One methodology and one register for security and AI risk, with AI system impact assessments (ISO 42001) feeding it. Keep per-framework views as filters on the one register, never as separate documents that drift apart.
Evaluate every risk against the Phase 0 appetite and record the treatment: mitigate, accept, transfer or avoid. This is where the early work pays off, because acceptance decisions stop being arbitrary and the same decision serves every framework.
Assess critical suppliers once against the combined requirements: CPS 230 material service providers, CPS 234 third-party controls, ISO 27001 supplier controls, and for AI providers, where prompts and outputs are kept and whether your data trains their models (ISM-2103).
Give each AI agent its own identity (ISM-2133), limit it to the tools and data it needs (ISM-2156, ISM-2157), require a person's approval for high-impact actions (ISM-2113) and log its activity centrally (ISM-2159). These satisfy ASD Essentials, the ISM and ISO 42001's operational controls together.
One intake, one severity ladder and one escalation path, whatever triggered the incident. Build each notification obligation into it: APRA within 72 hours for a material information security incident under CPS 234, the OAIC for eligible data breaches, and contract and government reporting duties.
Key risk and performance indicators a board will read, and a RACI naming who owns each control and process. Map each metric to the frameworks it evidences, so one dashboard serves every audit.
Phase 3: Certify, operate and evolve
Build the programme into procurement, the software development lifecycle, AI use-case approval and acquisitions, with role-based training, so it survives the people who designed it.
Decide what to certify, attest or self-assess, and in what order, based on the Phase 0 drivers. A common path: ISO 27001 first, ISO 42001 as an extension of the same management system, ASD Essentials assessed against the same controls, and SOC 2 or IRAP where a buyer requires it.
Tag each piece of evidence with every control it supports across the crosswalk, and collect continuously rather than before each audit. Overlapping audits stop being separate projects.
One report on the shared scale covering security and AI, with status against each framework as a view rather than a separate paper. This is also how a board evidences its accountability for AI (GOV-08).
Run internal audits and management reviews, revisit drivers and appetite with the steering committee, and track changes to the frameworks themselves, such as the Essential Eight giving way to ASD Essentials and new ISM releases.
An example project plan
What the checklist looks like as a schedule: an integrated ISO 27001, ISO 42001 and ASD Essentials programme, with an optional SOC 2 Type II report, for a mid-sized organisation, from mobilisation to certification in about ten months, as a work breakdown structure (WBS). Your timeline depends on scope and how much is already in place; treat this as a shape to adapt, not a promise.
Integrated programme: example schedule
Task Workstream rollup MilestoneWeeks are counted from mobilisation. Workstreams overlap on purpose: policy and risk work starts before the gap assessment ends, and controls go in while evidence is being collected.
Certification decision: what to certify, and in what order
Decide after the gap assessment. The answer is usually dictated by who is asking for the evidence.
Need someone to run it?
Running several frameworks as one programme is easier with someone who has done it before. Tell us which frameworks you need and we will reply within two business days, and can introduce you to an independent consultant who runs integrated ISO 27001, ISO 42001, ASD Essentials and APRA programmes.
CISO AI may receive a fee from providers we introduce you to. Your details are used only for this enquiry; see our privacy policy.
Questions
Can one programme cover ISO 27001 and ISO 42001?
Yes. ISO 42001 uses the same management system structure as ISO 27001 (context, leadership, planning, support, operation, performance evaluation, improvement), so an organisation certified to ISO 27001 can extend the same system to AI rather than build a second one. Many certification bodies audit the two together.
Which should come first, ISO 27001 or ISO 42001?
Usually ISO 27001, because most of the governance, risk and supplier machinery ISO 42001 needs is built there first. Lead with ISO 42001 only when AI is the product and the buyers asking for assurance are asking about AI specifically.
Does ASD Essentials replace ISO 27001?
No. ASD Essentials (the Essential Eight's successor) sets priority security outcomes and actions, and is not a certifiable management system. ISO 27001 certifies how you manage security. They fit together: the Essentials actions become controls inside the ISO 27001 system, assessed against ASD's effectiveness ratings.
Can SOC 2 and ISO 27001 share the same controls?
Yes. The Trust Services Criteria and ISO 27001's Annex A overlap heavily, so one control set mapped to both serves both audits. The difference is evidence: SOC 2 Type II tests that controls operated across an observation period, so evidence must be collected continuously rather than gathered before an audit.
Will an auditor accept a crosswalk?
Auditors assess each standard on its own terms, so a crosswalk does not replace evidence. What it does is let one control, one policy and one piece of evidence be presented to several audits, which is where most of the saving comes from.
We are APRA-regulated. Where does this leave CPS 234 and CPS 230?
They are obligations, not options, so they sit on your driver list from Phase 0 and their requirements map into the same control set. ISO 27001 is a common way to structure the programme that meets them. See our APRA CPS guide for how the standards divide the territory.
Where to go deeper
ASD Essentials and AI: the Essential Eight's replacement, principle by principle, mapped to the ISM AI controls.
ISO 42001: what the AI management system standard requires and how certification works.
APRA CPS 220, 230 and 234: how the prudential standards divide risk, resilience and security.