Briefing
Reference 8 · Primer · About 6 minutes
Download the MP3 · Leonard, an AI voice, reads a script written from the sources below.
Written from public sources and checked on 10 October 2026. The standards are paraphrased, not reproduced. Check the current text and your certifier's transition dates before relying on them.
The standards and frameworks that describe how to run AI governance as a system, and the practical work each one calls for.
These are Agile Insights AI Strategy & Governance offerings. Some are delivered with partners.
Sign up for our podcast at agile-insights.com.au.
Leonard Most organisations that want to govern artificial intelligence already hold two of the standards they need. They just do not know it. They are I S O, I E C, 27001 for information security, and 27701 for privacy. By the end of this episode you will know what they certify, what changed recently, and how they give you a head start on AI governance. This is our reference series, as at October 2026.
Leonard First, what they are. Both are published by the International Organization for Standardization and the International Electrotechnical Commission. Both are management system standards, which means they certify how you run something, not one product. Both are voluntary and certifiable. An accredited outside auditor checks you, and issues a certificate. 27001 covers information security. 27701 covers the management of personal information, which is privacy.
Leonard Now the problem they solve. Every customer questionnaire asks the same two things. Is our data safe with you, and do you handle personal information properly? A confident answer from the sales team is not evidence. A certificate from an independent auditor is. Think of a hospital group choosing a software supplier. It will usually ask for both before it signs.
Leonard Who uses them? Chief information security officers and privacy officers run them. Boards and audit committees rely on them. Suppliers sell with them, so they turn up in tenders, in vendor due diligence, and in regulator conversations after a breach.
Leonard What is in them? Four ideas carry the pair. One: both are risk based. You identify what could go wrong, then choose controls. For 27001, the reference list has 93 controls in four themes: organisational, people, physical and technological. Two: you record which controls you apply, and why, in a Statement of Applicability. Three: leaders own it. There is a policy, there are named roles, and there is a regular management review. Four: it must keep working. Internal audits, corrective action, and yearly surveillance audits are part of the deal. 27701 adds privacy duties, with separate controls for organisations that decide how personal information is used, and for those who process it on their behalf. Reported counts put it at about 78 controls.
Leonard There is a third standard people miss. I S O 31700 is about privacy by design for consumer products and services. 27701 manages privacy as a system. 31700 builds it into the product at design time. If a client says privacy by design, reach for 31700 as well.
Leonard Now the AI connection. The artificial intelligence management system standard, I S O, I E C, 42001, was built on the same skeleton as 27001 and 27701. That skeleton is called the Harmonised Structure. Context, leadership, planning, support, operation, evaluation and improvement are written once. So one risk method, one internal audit programme and one management review can serve all three. Only the lists of controls differ. For an organisation already certified, AI governance is an extension, not a fresh start. Be careful with sales claims about how much time this saves. The often quoted figure of half is an industry estimate.
Leonard AI also adds risks these standards were not written around. Security: training data can be poisoned, models can be stolen or manipulated, and prompts can be used to extract information. Privacy: personal information can sit inside training data, outputs can reveal it, and automated decisions about people need explaining. Neither standard tests whether a model is fair or accurate. That is what 42001 adds. Two more standards, one on AI security and one on AI privacy, are in the final stages, but I could not confirm either is published yet.
Leonard Getting started takes three steps. First, ask what you already certify, and what its scope covers. Second, run a gap analysis from those certificates to the AI standard, rather than starting a separate programme. Third, extend your risk method and your register of systems to include AI, with named owners. To learn more, start with the ISO catalogue pages, the Standards Australia store, and the privacy transition notice from the United Kingdom Accreditation Service. The show notes link to them.
Leonard Now currency, as at October 2026. For 27001, the current edition is the 2022 one. A 2024 amendment added climate change as an issue to consider. The deadline to move from the 2013 edition passed on 31 October 2025. Australia and New Zealand adopted the 2022 edition jointly. For 27701, a new edition was published on 14 October 2025. It stands alone, so you no longer need 27001 beneath it. The United Kingdom Accreditation Service has set 31 October 2028 for certification bodies to move all existing clients. Other accreditors may differ, so ask your certifier. The Australian edition still follows the 2019 text.
Leonard So, in short: security and privacy certificates are your fastest route into AI governance. If you'd like to know more, sign up for our podcast at agile-insights.com.au.
Every episode tagged #security
All of the reference series (16) · Follow AI Governance Reference in Apple Podcasts or any podcast app
Everything from Agile Insights: briefings, Making You Think and the reference series