The EU AI Act may affect your business even if you are based outside the EU. Start by identifying the AI you develop, supply or use, then match your compliance work to your role.
Importantly, you do not need the same controls for every tool. Start with a focused inventory, clear ownership and a record of why you made each decision.
What is the position today?
The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024. Its general application date was 2 August 2026, but some rules have later deadlines. The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026. Its amendments are law, not proposals, so plan against the amended rules. Later high-risk deadlines give you time to prepare; they do not pause rules already in effect.
Why your role and intended use matter
Your responsibilities depend on your role. Using third-party AI often makes you a deployer, whereas developing or commissioning AI under your own name may make you a provider. Supply-chain roles and certain changes to a system can also affect your obligations.
Buying a tool is not the end of the assessment. Ask what the supplier's assurance covers, which uses it supports and when changes would trigger a fresh review.
How you use a system matters too. The Act distinguishes prohibited practices, high-risk systems and uses that trigger transparency duties. Some Annex III uses qualify for a limited exception, but systems that profile individuals remain high risk.
For example, a tool that finds human resources policies needs a different assessment from one that ranks job applicants.
If you provide a general-purpose AI model, separate documentation, copyright-policy and training-content-summary duties apply, with extra requirements for systemic-risk models. However, buying an application built on a model is a different role.
Why organisations outside the EU should pay attention
The Act reaches providers supplying the EU market and can cover providers and deployers outside the EU where system outputs are used there, subject to scope limitations and exclusions.
If you are based in the UK, map your EU connections rather than relying on where your headquarters sits. Candidate assessments produced abroad for an EU recruitment process, for example, warrant a scope review.
For this, it is important to bring legal, procurement, technology and operations together early. Record where the system is supplied and its outputs are used, and who relies on them. Local requirements may also have to be checked separately.
Employers: prioritise recruitment and worker-management tools
Employment-related high-risk categories include recruitment, candidate evaluation, promotion and termination decisions, certain task allocation and worker monitoring. Workplace emotion inference based on biometric data is generally prohibited, except for medical or safety reasons.
Start with tools that assess people or influence important employment decisions. Ask what they infer, how their recommendations affect outcomes and whether a human can genuinely challenge them. Notably, an approval click alone is not meaningful oversight.
When high-risk deployer rules apply, they cover instructions, human oversight, monitoring, logs and incident escalation. Employers must inform affected workers and their representatives before deploying a high-risk system in the workplace. Check any applicable national information and consultation rules too.
Transparency and AI literacy need attention now
Transparency duties distinguish provider marking from deployer disclosure. They cover specified AI interactions and generated or manipulated content, including deepfakes, with exceptions.
The 2 December 2026 marking transition concerns qualifying systems marketed before 2 August 2026; it does not postpone every transparency duty.
Make sure that you are checking customer-facing tools and publishing workflows. Decide which notices or disclosures you need, who owns them and how to update them when features change.
The amended literacy rule calls for measures that support staff and other operators in developing AI literacy, tailored to their experience and how they use AI. It does not require you to guarantee a particular level of knowledge for each person.
Make training practical: help recruiters question candidate rankings, communications teams spot disclosure issues and system owners report problems or pause use.
Enforcement: substantial exposure, not automatic maximum fines
Non-compliance has consequences. Prohibited-practice breaches can attract fines of up to €35 million or 7% of an undertaking's preceding financial year's worldwide annual turnover, whichever is higher. For specified other breaches, including transparency failures, the ceilings are €15 million or 3%.
Five practical steps to take now
1. Separate current duties from preparation work
For each system, separate current duties from later deadlines. Assign an owner, record the legal basis and check whether its introduction date, design or a transition changes the timetable.
2. Build a focused AI inventory
Record each system's purpose, supplier, owner, users, markets and EU connections. Identify your role, prioritise uses affecting people or safety, and flag uncertain classifications for review.
3. Ask suppliers for usable evidence
Ask for intended-use documents, limitations, oversight instructions, available records and change-management details. Check that your contract supports monitoring and incident response, and clarify what any compliance claim covers.
4. Make oversight operational
Name the people who approve deployments, review changes and investigate concerns. Make sure they can challenge outputs, and give staff a clear way to report problems and stop unsafe use.
5. Maintain a legally grounded monitoring plan
Separate law from proposals, guidance and voluntary codes. Record the source and reason for each deadline, and revisit your assessment if a system's purpose, market or design changes.
The takeaway
Know what AI you use, where you use it and who is responsible. That is a better starting point than a single deadline or a supplier's blanket assurance.
Not sure where to start? Ethiqs Legal can help you work through the AI governance and EU compliance questions that matter to your business. Book a free 20-minute call with the Ethiqs team.