EU AI Act: what it means for firms outside the EU with EU clients
By the ChamberOne team 5 minute read
You practise in Colombo, London, Singapore or Chicago. You have a client in Europe, and their procurement team has just sent a questionnaire that asks how you use artificial intelligence and whether you comply with the EU AI Act. Do you? The honest answer is that it depends, and that the first job is to understand which part of the regulation could touch you at all. This article is general information, not advice. Take specific advice on the current text and any guidance, because both are still developing.
Key points
- A firm outside the EU may meet the Act through client contracts even where it does not apply directly.
- Buying and using a tool generally keeps a firm on the deployer side. Building and offering one to others carries more responsibility.
- Typical law firm tasks are unlikely to be in the highest risk categories, but check work that evaluates people or supports public bodies.
- An inventory, a short policy and training records answer most client questionnaires.
Why a firm outside the EU might care
European legislation can reach beyond Europe's borders. Whether it applies to a given firm depends on where the AI system is used, where its output is used and the role the firm plays. A firm with no European office and a handful of European clients is not in the same position as one that markets to European consumers or runs staff there.
Even where the law does not apply directly, clients may ask you to behave as if it did. Large organisations push their obligations down the supply chain through contract terms and questionnaires. For many firms, the first contact with the Act will be a clause in a services agreement.
Providers and deployers: which one are you?
The regulation distinguishes between those who build or place an AI system on the market and those who use it. A law firm that buys a tool and uses it for its own work is generally a user of that tool, which the Act calls a deployer. A firm that builds its own tool and offers it to others is closer to a provider, with heavier responsibilities.
This is one more reason to think carefully before building. Buying a product keeps the firm on the user side of the line for most purposes. Confirm this with specialist advice rather than assuming it, particularly if you customise a tool substantially or put your own name on it.
- Buys a tool and uses it for its own work
- Buying keeps the firm on the user side for most purposes
- Builds its own tool and offers it to others
- Carries heavier responsibilities
- Customising substantially or using your own name needs specialist advice
Where law firm use is likely to sit
The Act sorts uses into categories according to risk, with the strictest rules reserved for uses that threaten people's safety or fundamental rights. Typical law firm tasks, such as summarising documents, drafting letters, extracting clauses and searching internal knowledge, are unlikely to fall into the highest categories. They are more likely to attract light obligations, such as being open about AI use in some situations.
That is a general expectation, not a clearance. Some legal work comes close to areas the Act treats more seriously, such as assisting a court in applying the law, or tools that evaluate individuals for employment decisions. A firm advising an HR client on a screening tool, or helping a public body, should look at the details of what the system does. Do not assume that because you are a law firm, the Act is irrelevant.
Staff understanding of AI
One theme that recurs in discussions of the Act is that people who use AI systems should understand them well enough to use them sensibly. Whether or not the specific provision binds your firm, the expectation is a good one. A firm that cannot show that its staff have been trained, know the limits of the tools and understand what must be checked is exposed under almost any regulator's thinking.
Keep a simple record of who has been trained, when and on what. It costs little and answers a question clients are likely to ask.
What a client questionnaire will probably ask
Expect questions along these lines. Preparing short, honest answers in advance saves time across every client.
- Which AI tools does the firm use on our matters, and for what purposes?
- Is our data used to train any model?
- Where is data processed, and which providers receive it?
- Is there human review of AI output before it reaches us?
- Is AI use recorded, and can you show us the record?
- Who in the firm is responsible for AI governance?
- Can we ask you not to use AI on our matters?
A practical preparation plan
- 1
Make an inventory of every AI tool in use, including those individuals brought in themselves.
- 2
For each, note the provider, what data it receives and whether there is a record of use.
- 3
Classify each use in plain terms: internal efficiency, client-facing output or anything that evaluates people.
- 4
Write a short policy: approved tools, prohibited uses, review requirements and who to ask.
- 5
Prepare a one-page statement you can adapt for client questionnaires.
- 6
Agree a process for a client who opts out of AI on their matters, and make sure staff can apply it.
Contracts and data transfers
AI rules rarely stand alone. They sit alongside data protection law, confidentiality duties and the terms of your agreement with the client. Look at what the contract says about subcontractors, because an AI provider may count as one. Look at whether personal data of people in Europe leaves Europe when you use the tool. These questions existed before AI and still matter more than the AI label.
Keeping a sense of proportion
It is easy to treat a new regulation as a crisis. For most firms outside Europe the sensible response is modest and mostly consists of good habits they should have anyway: know what tools are in use, keep records, train people and be able to explain it all to a client in plain words. Do that, and a questionnaire becomes an afternoon's work rather than a project.
What this article cannot settle
The Act is detailed, and guidance on how it applies is still being issued. Dates of application and obligations are staged, and secondary materials change. Treat anything you read, including this, as a starting point for a conversation with a specialist who has the current text in front of them. If a client asks for assurance, answer with facts about your practice, such as the tools, the controls and the training, and avoid claiming a compliance status you have not had checked.
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
Generative AI guidance for lawyers: how six markets compare
How to find what regulators, courts and bar bodies have said in six markets, and the common themes across them.
5 minute readAustralia: court practice notes on generative AI and what they require
Courts in Australia have issued practice notes on generative AI. Read the one for your court, every time.
4 minute readAustralia: the Privacy Act and AI in law firms
AI tools process personal information. Australian firms should map where that information goes before adopting them.
4 minute read