Briefing
Reference 2 · Primer · About 5 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 law is paraphrased, not reproduced. Check the current text before relying on it.
The rules that apply to AI in Australia and for Australian organisations with European exposure, and the practical work each one calls for.
These are Agile Insights AI Strategy & Governance offerings. Which of them you need depends on where your organisation starts.
Sign up for our podcast at agile-insights.com.au.
Leonard The European Union's Artificial Intelligence Act is the first broad AI law in the world, and it can reach an Australian company with no office in Europe. By the end of this episode, you will know who it covers, what is already enforceable today, and which dates have moved.
Leonard This is a primer, as at October 2026. First, what it is. The Act is Regulation 2024/1689, a law of the European Union. It applies directly in every member state, and it is mandatory. It is not a standard you choose to adopt. It sorts AI by the risk it poses to health, safety and fundamental rights, and the higher the risk, the heavier the duty.
Leonard The problem it solves is trust at scale. Imagine a recruitment tool that screens thousands of applicants and quietly disadvantages one group. No individual sees the harm, yet the organisation carries it. The Act makes the people who build and use such systems answerable, with fines of up to 35 million euros or seven percent of worldwide turnover for the worst breaches.
Leonard Who uses it? Boards and risk leads, when a European customer asks for assurance. Procurement teams, when they buy AI from a vendor. Legal counsel, when a product launches into Europe. The Act has roles. A provider builds or brands the system. A deployer uses it in a professional setting. Most Australian organisations are deployers, but some are providers without realising it.
Leonard Now, what is in it. Four ideas carry the Act. The first is four risk tiers. Unacceptable risk is banned outright: social scoring, manipulative systems, and emotion recognition at work or school. High risk is heavily regulated, and covers uses such as hiring, education, credit, essential services and law enforcement. Limited risk carries transparency duties only. Minimal risk, which is most AI, has no obligations.
Leonard The second idea is the high risk package. Providers need risk management, quality data, technical records, logging, human oversight, and a conformity assessment before the product reaches the market. Deployers have lighter but real duties, such as using the system as instructed and keeping humans in charge. The third idea is general purpose AI models, meaning the large foundation models behind chatbots. Their providers must document the model, respect copyright, and summarise training content. The most powerful models face extra testing and incident reporting.
Leonard The fourth idea is reach. The Act applies if you place a system on the European market, or if your system's output is used in the EU, even from Sydney. So an Australian firm with European clients or staff may be caught.
Leonard How does it connect to the rest of your stack? The Act is the law. Standards such as I S O, I E C, 42001, the AI management system standard, are how you organise to meet it. They help, but they do not replace legal compliance. Underneath sit your data foundations, because high risk systems need well governed data. Australia has no equivalent AI statute, so for many firms this is the strictest rule they will meet.
Leonard Getting started takes three steps. Build an inventory of the AI systems you build, buy and use. Tier each one honestly, and confirm your role and your European exposure. Then run a gap analysis for the few that are high risk or face Europe. Your primary sources are the European Commission's AI Act pages and the official text on EUR-Lex, linked in the show notes.
Leonard Now the currency check, as at 10 October 2026. Already applying since 2 February 2025: the bans, and a duty to support AI literacy. Since 2 August 2025: the rules for general purpose models. Since 2 August 2026: the transparency rules in Article 50, and the Commission's enforcement powers over model providers. Article 50 means people must be told when they deal with a chatbot, and synthetic content must be marked and labelled. Providers of generative systems already on the market have until 2 December 2026 to add the machine readable marking.
Leonard And the delay. A law called the Digital Omnibus on AI, Regulation 2026/1744, was published on 24 July 2026 and is in force. It has moved the high risk dates. Stand alone high risk uses now apply from 2 December 2027. High risk AI built into regulated products, such as medical devices, applies from 2 August 2028. It also adds two new bans, on AI that creates non consensual intimate images and child abuse material, from December 2026. So the delay is adopted, not proposed. But it is a reason to plan, not to stop. Check the current text before relying on any date.
Leonard To recap: a risk based European law that can reach you from Australia, partly live now, with the heaviest duties arriving in 2027 and 2028. If you'd like to know more, sign up for our podcast at agile-insights.com.au.
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