ISO 42001 for Australian companies: what it is and whether you need it
The AI management system standard Australian buyers are starting to ask about, how it lines up with the local frameworks, and a ten question gap check.
ISO/IEC 42001 is the international standard for managing how a business builds and uses AI. It was published in December 2023 and adopted in Australia as AS ISO/IEC 42001:2023 in February 2024. An accredited auditor can certify you against it, the way ISO 27001 works. Most Australian companies only need the certificate when a customer or tender asks for one.
Everyone else can work towards it first, using the voluntary Australian guidance most of it already overlaps with. This guide covers what the standard asks for, who is asking Australian suppliers about it, how it lines up with the Australian frameworks, and a gap checklist you can run this week. For the wider picture, start with our guide to AI governance in Australia.
What is ISO 42001, in plain terms?
It describes an AI management system: the policy, people, records and routines a business uses to keep its AI use under control and improving. It does not tell you which model to use or set a test score to hit. It tells you to decide what good looks like for your business, write it down, do it, check it and fix what the checks find.
The main body follows the same clause structure as ISO 27001 and ISO 9001, so a business already certified to either recognises the shape. Clauses 4 to 10 cover context, leadership, planning, support, operation, performance evaluation and improvement. Annex A lists 38 controls under nine objectives, from AI policy and roles to data, impact assessment, system life cycle, use of AI systems and third-party relationships. Annex B is the guidance on how to implement them.
Four pieces of work show up in almost every audit:
- Scope and role. Which AI the system covers, and whether you are a provider, a producer or a customer of AI. Most Australian mid-market businesses are customers: they use Claude, Copilot or HubSpot's AI in their own work.
- An AI policy signed by leadership, with named people responsible for each AI system.
- Risk and impact assessment. What each use could do to the business and to the people it touches, and what you will do about it.
- A Statement of Applicability listing which Annex A controls you apply, which you exclude, and why.
Simplified. The standard's clauses 4 to 10 follow the same plan, do, check, act pattern as ISO 27001.
Who in Australia is asking for it?
Mostly larger buyers with their own security or risk team: government and the businesses that supply it, health providers, professional services groups and companies in the defence supply chain. The request usually arrives as a supplier security questionnaire, sent once you are on the shortlist and before the contract is signed.
The questionnaires we answered this year did not name ISO 42001. What they asked for was the work the standard describes. One buyer's IT security team wanted, in a single email: AI governance documentation, AI security and data flow documentation, the versions and processing regions of the models we use, data processing agreements with our AI providers, controls for prompt injection and data leakage, logging that would catch misuse, an incident response plan, and any certifications such as ISO 27001 or SOC 2.
That is close to a reading list of ISO 42001's Annex A. A business that has done the work can answer every line, certificate or not. One that has not will struggle, whatever logo it has. Our post on the AI security questionnaire your buyer's CTO is about to send goes through the questions one by one and what a good answer looks like.
Do you need ISO 42001 certification?
Only if someone you sell to requires it. Certification is evidence for other people. The work underneath it is what protects you. Use this table to decide.
| Your situation | What to do |
|---|---|
| A tender or contract lists ISO 42001 as mandatory | Certify. Start with a gap assessment and plan for a full audit cycle before the deadline. |
| You sell a product with AI built in to large organisations | Strong case for certifying. You are a provider of AI, and procurement teams will keep asking. |
| Questionnaires ask about AI governance but not the standard | Do the work first. Answer with your policy, register of uses, impact notes and logs. Revisit certification if buyers start naming it. |
| You are APRA regulated, or supply a business that is | Meet CPS 230 first, because it is law for regulated entities. Working to ISO 42001 helps you evidence how AI suppliers are managed. |
| You use AI tools internally and sell nothing with AI in it | Work to the 6 practices in Australia's Guidance for AI Adoption. Certification rarely pays its way here. |
Check your vendors too. Anthropic, which makes Claude, lists ISO/IEC 42001:2023 among its certifications for its commercial products, alongside ISO 27001 and SOC 2. A vendor's certificate covers how the vendor runs its AI. It does not cover how your business uses it, so you still need your own policy, owners and register.
How does ISO 42001 map to the Australian AI frameworks?
Australia has no AI Act in force. What a business here faces is a mix of voluntary guidance, security controls and two pieces of law with real dates. ISO 42001 overlaps with each one differently.
| Australian framework | What it is | Where ISO 42001 fits |
|---|---|---|
| Guidance for AI Adoption (the 6 practices) | Voluntary guidance from the National AI Centre, published October 2025. It evolves the 2024 Voluntary AI Safety Standard. | Broadly in line. Accountability, impact assessment, risk management, transparency, testing and human control all have matching clauses or Annex A controls. |
| Essential Eight | Eight cyber security controls from the Australian Signals Directorate. Nothing in it is specific to AI. | Does not replace it. ISO 42001 assumes your information security is handled elsewhere, by the Essential Eight, ISO 27001 or both. |
| Privacy Act and the APPs | Law. From 10 December 2026, privacy policies must explain decisions made by a computer program that significantly affect people. | Certification does not satisfy the law, but its impact assessments are how you find the decisions you have to disclose. |
| APRA CPS 230 | Law for banks, insurers and super trustees, covering operational risk and material service providers. | Supplier records and AI risk assessments under ISO 42001 give regulated buyers the evidence they ask suppliers for. |
| ISO 27001 | The international standard for information security management. | Same clause structure, so the two can run as one integrated management system with one internal audit. |
Each framework has its own post. Our post on which Australian AI framework applies to you explains whether the 2024 Voluntary AI Safety Standard or the 6 practices fits your business. Does the Essential Eight cover AI? covers what the eight controls leave open. How to secure AI sums up what the ACSC says. The Privacy Act change on 10 December 2026 sets out the disclosure duty, and is your AI vendor a material service provider under CPS 230? covers APRA.
What does an ISO 42001 gap checklist look like?
Run through these ten questions with whoever owns AI in your business. Each "no" is a gap. If most come back "yes", you already do most of what the standard asks, and certification is mostly a matter of keeping records an auditor can read.
- Scope. Have you written down which AI the system covers, and whether you are a provider, producer or customer of AI?
- Policy. Is there a short AI policy, signed by leadership, that staff have actually read? Our one-page AI policy is a starting point.
- Owners. Does every AI system have one named person who knows they own it?
- Register of uses. Do you list what each tool is used for, and not just which tools you pay for? The same tool carries different risk in marketing copy and in hiring.
- Impact assessment. For each use that touches customers, staff or decisions about people, is there a written note on what could go wrong and for whom?
- Statement of Applicability. Have you gone through Annex A and recorded which controls apply and why the rest do not?
- Suppliers. Do you hold each AI vendor's certificates and data processing terms, and know where they process data? Sending personal information to a model hosted overseas engages APP 8 on cross-border disclosure.
- Testing and logs. Is there a test set you rerun when a model or prompt changes, and a log someone reads?
- Human control. Can a person override or switch off each system, and has anyone actually done it?
- Review. Is AI on the agenda of an internal audit and a management review at least once a year?
Our AI governance framework template turns most of this list into documents you can fill in.
How does ISO 42001 certification work in Australia?
The process matches other ISO management system standards. You close the gaps, run the system long enough to produce records, then book an accredited certification body. In Australia and New Zealand, certification bodies are accredited by JAS-ANZ, so check that any auditor you hire holds accreditation for ISO/IEC 42001 specifically.
- Gap assessment against the clauses and Annex A, done internally or with an adviser.
- Implementation of the policy, assessments, controls and records.
- Internal audit and management review, so the auditor sees the system checking itself.
- Stage 1 audit, where the certification body reviews your documents and readiness.
- Stage 2 audit, where they test that the system works in practice.
- Certificate, usually valid for three years, with surveillance audits in between.
The cost is the auditor's fees plus your own team's time, and the time is usually the bigger number. Only an accredited body can certify you, so get quotes from two or three of them and ask each how many audit days they expect for your size and scope.
Where does Neighbourhood stand?
We would rather tell you plainly. Neighbourhood is not certified to ISO 42001, and we do not hold ISO 27001 or SOC 2 either. We are working towards Essential Eight Maturity Level 1, with an external penetration test planned for the first half of 2027 and SOC 2 Type I readiness and an ISO 27001 gap assessment planned for 2027. Anthropic, the model provider most of our AI work runs on, holds ISO 42001 and ISO 27001.
Everything above, plus our subprocessors and the document bundle we send buyers, is on our trust centre. If you are filling in your own questionnaire and want to see how a supplier answers one honestly, start there.
Frequently asked questions
Is ISO 42001 mandatory in Australia? No. No Australian law requires it. It becomes mandatory for you only when a customer contract or tender makes it a condition of the work.
What is the difference between ISO 42001 and ISO 27001? ISO 27001 manages information security: keeping data confidential, accurate and available. ISO 42001 manages AI: how AI systems are chosen, assessed, tested and overseen. They share a clause structure, so many businesses run them together.
Is AS ISO/IEC 42001:2023 different from ISO/IEC 42001? No. Standards Australia adopted the international standard in February 2024 under the AS ISO/IEC label, so it is the same standard, published locally.
If our AI vendor is ISO 42001 certified, are we covered? No. The vendor's certificate covers how the vendor builds and runs its AI. How your business uses that AI, what data goes in and who checks the output is still yours to manage.
Who certifies ISO 42001 in Australia? Certification bodies accredited by JAS-ANZ, the accreditation body for Australia and New Zealand. Ask any auditor to show accreditation for ISO/IEC 42001 itself, not only for ISO 27001.
Read next
- AI governance in Australia: the 2026 guide
- The AI security questionnaire your buyer's CTO is about to send
- An AI governance framework template for Australian businesses
- The Essential Eight explained, and what it means for a business using AI
Sources
- Standards Australia adopts the international standard for AI Management System, AS ISO/IEC 42001:2023, Standards Australia, 16 February 2024.
- ISO/IEC 42001:2023, Information technology, Artificial intelligence, Management system, ISO, for the clause structure and Annex A.
- What certifications has Anthropic obtained?, Anthropic Privacy Center, for Anthropic's ISO/IEC 42001:2023 and ISO 27001 certifications.
- Guidance for AI Adoption, National AI Centre, Department of Industry, Science and Resources, 21 October 2025.
- APP guidelines, chapter 8: cross-border disclosure of personal information, OAIC.
- Privacy and Other Legislation Amendment Act 2024, for the automated decision-making disclosure obligation commencing 10 December 2026.
- Trust centre, Neighbourhood, for our own certifications and roadmap.
Neighbourhood builds AI systems on Claude for Australian businesses, and helps them put the governance around those systems that buyers and regulators ask about, as part of our AI engineering work. We have worked with 300+ businesses since 2015.
Got an AI questionnaire you are not sure how to answer?
Tell Matt what the buyer is asking for. He will talk you through where you stand and what to fix first.