EU AI Act: what it means for Australian online stores
Read an EU AI Act summary for online stores: when it applies outside the EU, what it asks of chatbots and AI content, and the dates to plan for.
- Published
- Last updated
The EU AI Act regulates AI according to its use and risks, and it can cover an Australian store when its AI output is used in the EU. For a typical shopping chatbot, the immediate task is making the AI interaction clear to customers. Using AI to assess job applicants or creditworthiness raises different obligations.
The chatbot transparency rules already apply. The later deadlines for high-risk systems don’t postpone those duties. This guide gives general information for store owners. Get legal advice for a specific deployment or a disputed classification.
EU AI Act summary
The EU AI Act covers providers that supply AI systems and deployers that use them under their authority. The European Commission’s overview explains four risk levels:
| Risk level | What it means | Example relevant to a retailer |
|---|---|---|
| Unacceptable | Specified practices are prohibited | Prohibited social scoring or harmful manipulation |
| High | Listed uses and certain regulated-product systems face detailed requirements | AI that screens job applicants |
| Limited | Certain systems have transparency duties | A chatbot that talks to customers |
| Minimal or no risk | The use has no specific risk-based requirements under the Act | An ordinary spam filter |
The four levels are an overview, not four certificates a vendor can buy. General-purpose AI models have their own rules. A store using a supplier’s model through an app doesn’t automatically become the model’s provider.
The Act takes effect in stages. AI literacy and the original prohibitions began applying in 2025, followed by rules for general-purpose AI model providers. Chatbot transparency duties now apply. High-risk duties have later application boundaries, with separate stages for listed high-risk uses and systems embedded in regulated products.
The Digital Omnibus changed those high-risk deadlines. For the exact boundaries and transitional rules, use the Commission’s implementation timeline and Article 113 on EUR-Lex. Providers of some previously marketed systems have a separate transition for machine-readable content marking. That transition doesn’t extend the customer-facing chatbot disclosure rule.
Does it apply to an Australian store?
Yes, the Act can apply to an Australian store when output from its AI system is used in the EU. Having your business and servers in Australia doesn’t by itself put the use outside the Act.
Article 2 on EUR-Lex includes providers and deployers outside the EU where AI output is used there. It also covers providers placing AI systems on the EU market, wherever the provider is established.
For example, an Australian store’s chatbot gives product recommendations to customers shopping from France. That is a practical reason to assess the EU rules. Merely knowing that your website can load overseas doesn’t settle every question about territorial scope.
A store using a vendor’s chatbot is usually a deployer. A business that develops a system, or has one developed and supplies it under its own name, can be a provider. The Commission’s transparency guidance explains those roles. Establish the roles in your deployment before assigning tasks between your team and the vendor.
EU AI Act compliance for chatbots
EU AI Act compliance for a shopping chatbot starts with telling people they’re interacting with AI. Article 50 on EUR-Lex places the design duty on the provider, unless the AI interaction is obvious in context. The information must be clear and accessible by the first interaction.
As the store deploying the chat, inspect the customer experience. A visible introduction such as “You’re chatting with our AI shopping assistant” makes the interaction clear. A chatbot that uses a person’s name and photo without disclosure creates a different impression. Confirm the notice remains visible on mobile and after changes to the widget.
Content rules distinguish what the AI provider must build from what your store must disclose:
- Providers of generative systems must support machine-readable marking and detection of synthetic output, subject to the Act’s exceptions.
- Deployers must disclose AI-generated or manipulated images, audio or video that constitute deepfakes. This can include realistic fabricated content that falsely appears authentic.
- AI-generated text published to inform the public on matters of public interest has a disclosure duty. The exception requires human review or editorial control and a person or organisation holding editorial responsibility.
These rules don’t create a blanket requirement to put a visible AI label on every product description. They also don’t make a visible label a substitute for the provider’s machine-readable marking duty. The official Article 50 explanation sets out the separate duties and exceptions.
For a hypothetical campaign, an AI-made video shows a real person endorsing a product they never endorsed. Assess the deepfake disclosure duty before publishing it. Labelling the clip doesn’t resolve consent or other legal requirements.
AI agents and audits under the EU AI Act
An AI agent’s purpose determines its obligations. Calling software an agent doesn’t automatically make it high-risk or exempt it from the Act. The Commission’s service desk explains that agents interacting with people or generating content can have transparency duties.
A shopping agent that recommends products and adds them to a basket ordinarily faces the chatbot transparency requirements. An AI tool that ranks job applicants needs a separate high-risk assessment. Check the intended purpose against Article 6 and the high-risk categories on EUR-Lex.
The Act doesn’t require every store chatbot to obtain an independent AI audit. For high-risk systems, Article 43 on EUR-Lex sets conformity assessment procedures. Some use internal control, while others require a notified body. A notified body is an organisation authorised to perform the relevant independent assessment.
Ask a vendor claiming to be “EU AI Act certified” which system and requirement the claim covers. A general management-system certificate doesn’t replace the applicable conformity assessment. Keep practical testing of the chatbot’s answers and actions in your purchasing process even when a formal high-risk assessment isn’t required.
AI literacy duty
The AI literacy duty requires providers and deployers to take measures that support staff understanding of the AI they operate. It began applying in 2025. Article 4 on EUR-Lex requires attention to staff experience and the context of use. It doesn’t require a guaranteed literacy level for each individual.
For your store, teach the people managing the chatbot what information it uses and which actions it can take. Show them how to recognise an incorrect answer, handle a customer escalation and stop an unsafe action. Give content staff separate guidance on synthetic media and disclosure.
Keep a record of the instruction and who received it. The Commission’s AI literacy answers explain that a particular training certificate isn’t required. The measures need to suit the work staff perform.
Check your use
Use the EU AI Act compliance checker to work through your store’s role and AI uses. Assess each use separately: customer chat, content creation and recruitment can lead to different results. Take that record to your vendor or legal adviser when a role or high-risk classification needs resolving.
Give your store its own shopping agent
Customers ask in their own words. Your agent finds, compares and checks out from your catalogue, right on your site.