Control over AI
Blog
AI governance 6 min read

Which AI tools may lawyers use?

The professional rules name no tools, they name core values. Which AI tools a lawyer may use is therefore a firm decision: a list with statuses, conditions and alternatives, and a route for the tool that is not on it yet.

Abstract sphere between translucent panels
Quick answer

Which AI tools a lawyer may use is not in the professional rules and not on the tool: it is a firm decision. The Netherlands Bar gives recommendations along the profession's core values: the lawyer knows the data flows and puts no confidential data into free tools. The firm translates that into a list with, per tool, a status (allowed, limited use, no decision yet, not allowed), for limited use the condition about case information, and for a rejected tool the alternative. A tool that is not on the list has the status no decision yet, and until that decision only public information goes in.

01

The professional rules name core values, not tools; the list is a firm decision

02

Every tool has one of four statuses, and a free and a business account are two rows

03

Limited use means for a lawyer: the condition says which case information may go in

04

A tool that is not on the list is no decision yet, neither a ban nor a free pass

05

A client can be stricter in their terms than the firm's list

A trainee asks her supervising partner whether she may use Claude to organise the structure of a pleading. He thinks for a moment. "The research database is fine, we have a licence for that. The rest, honestly, I do not know." She uses Claude that evening anyway, at home, on her own account.

Which AI tools a lawyer may use is not in the professional rules and not on the tool. It is a decision of the firm. A firm that does not take that decision is not a strict firm; it is a firm where every supervising partner works it out alone, and where the work drifts to private accounts.

Why is the answer not in the professional rules?

Because the professional rules are about core values, not products. A lawyer is independent, partisan, competent, honest and confidential, and the dean supervises those core values. The Netherlands Bar has published recommendations on AI in the legal profession organised along the same core values. They say how to assess a tool: know and control the data flows, put no confidential data into free tools, check what the tool produces, stay responsible for the advice yourself. They name no tool that is allowed and no tool that is not.

That is not a gap in the recommendations; it is the only way they can be correct. The same chatbot on a free account is something different from the same chatbot on a business licence with a processing agreement, and a contract analysis tool connected to the case management system is something different from the same tool without that connection. The bar can give the core values; the firm has to apply them per tool. How that works for organisations outside the legal profession is in which AI tools may employees use; for a firm, the duty of confidentiality is added as the heaviest condition.

What does a firm's list look like?

Per tool a status, and four are enough: allowed, limited use, no decision yet, not allowed. What they mean is in four statuses for AI tools. Limited use comes with the condition, not allowed with the alternative, and a free and a business account of the same tool are two rows. This is what it might look like at a mid-sized firm; the decisions are examples, because your firm will decide differently.

ToolStatusCondition or alternative
Legal research database on the firm's licenceAllowedNone; case information permitted, within the agreement with the vendor
Microsoft 365 Copilot on a work accountAllowedOnly in matters you hold permissions on yourself; nothing from walled-off matters
ChatGPT or Claude through the firm's business accountLimited useNo party names, case numbers, amounts or documents; abstract legal questions and your own text without client data are fine
ChatGPT or Claude through a free or personal accountNot allowedAlternative: the firm's business account
Transcription tool for meetingsLimited useInternal meetings only, participants know; no client meetings
Contract analysis tool first seen this monthNo decision yetPublic information only for now; assessment under way

The list does not live in a PDF. A lawyer opening a tool at a quarter past ten does not read a protocol; the decision should appear the moment the tool opens, with the condition attached and the alternative as the first button. This is what that looks like when the list sits at the point of use instead of in a document:

Search tool, vendor or categoryโ€ฆ
865 tools
ChatGPT๐Ÿ‡บ๐Ÿ‡ธ
HighAllowed
Claude๐Ÿ‡บ๐Ÿ‡ธ
MediumAllowed
DeepSeek๐Ÿ‡จ๐Ÿ‡ณnew
CriticalNot allowed
Perplexity๐Ÿ‡บ๐Ÿ‡ธnew
MediumNo decision yet
Mistral๐Ÿ‡ซ๐Ÿ‡ท
LowNo decision yet
Midjourney๐Ÿ‡บ๐Ÿ‡ธ
MediumNo decision yet

Every tool, scored for risk

865 AI tools, each scored Low to Critical, with nothing pre-approved or pre-blocked until someone decides.

More about AI Tools

What does limited use mean for a lawyer?

More than for most professions, because the condition is almost always about case information. "No party names, case numbers, amounts or documents" is concrete enough to follow. "Be careful with client data" is not.

There is a trap specific to legal work. A legal question can be traceable on its own. Whoever asks about terminating the lease of a cinema in a town with one cinema has named the client without typing a name. A matter that has been in the press, a small niche, an unusual combination of facts: the condition "no names" does not cover it, and the lawyer has to see it themselves. Which data together make someone identifiable is in what you can and cannot share with AI, and why the account is half the difference in free vs business AI accounts.

What do you do with a tool that is not on it?

It has the status no decision yet. That is neither a ban nor a free pass: for now only public information goes in, and the tool is reported to whoever maintains the list. Trying it is not a breach. Putting a case file in before the decision is made is.

A twenty-five-lawyer firm has a list of six rows. A first look at which AI services are opened from the firm's browsers shows that a contract analysis tool recommended at a conference is used by three corporate associates, and that a transcription tool someone uses for client meetings appears nowhere. Nobody did anything that was not allowed, because nothing had been decided. The contract analysis tool gets a row on no decision yet, the head of the practice group answers the questions to the vendor, and two weeks later it stands on limited use: for contracts of clients who have agreed to it, without a connection to the case management system. The transcription tool goes to not allowed, with the firm's dictation software as the alternative.

Such a route has three parts: who assesses, within how much time, and on the basis of which questions. Those questions are in approving an AI tool: what to ask the vendor. The decision itself lands in the firm's AI register.

What if the client has agreed something else?

Then the client comes first. Engagement letters and procurement terms increasingly say something about AI, from a duty to notify to a ban on external AI services on the matter. A tool that is allowed for the firm can be not allowed on that one matter. The firm's list is the floor; the agreement with the client can be stricter, never looser. Check it per matter and record it where the whole team on the file sees it.

BeeSensible sets a catalog of 865 AI tools against the tools opened from the firm, with no user id and no content. An administrator sees which tools still have no decision and records the firm's decision, with the condition or the alternative; that decision appears to the lawyer the moment they open the tool, and continuing is always possible. How that works for a firm is on the page for legal organisations and on the AI tools product page. The larger whole, from seeing what is used to showing the agreements are followed, is in AI governance for law firms and under the topic AI governance.

Next time the trainee asks, the answer is not "honestly, I do not know" but "check the list, and if it is not on there, that is its status". It is a smaller answer, and it keeps the pleading on the firm's account.

FAQ

Common questions

Is there a list of AI tools lawyers are allowed to use?

Not from the Netherlands Bar. The bar has published recommendations on AI along the profession's core values, and those say how to assess a tool, not which tool. The firm makes the list itself, per tool, with a status and a condition. Two firms can rightly decide differently about the same tool, because their practice, their clients and their accounts differ.

May I use ChatGPT as a lawyer?

That depends on the account and on what your firm has decided. A free or personal account runs under consumer terms, without a processing agreement, and input can be used for training unless you switch it off; no case information belongs there. A business tier held by the firm can be limited use, for example without party names and case numbers. Check your firm's list, and if the tool is not on it, that is its status.

What does limited use mean for case work?

That the tool may be used within one condition the firm has written down, usually about case information: anonymised text only, no party names, case numbers or amounts, or public sources only. Bear in mind that a legal question can itself be traceable if the matter has been in the press or the niche is small. When in doubt, ask the owner of the row.

What do I do with an AI tool that is not on the list?

Treat it as no decision yet: public information only, no case information, and report it to whoever maintains the list. A well-run firm has a short route for new tools: who assesses, within how much time, on the basis of which questions. Trying a tool is not a breach, as long as no case file goes in before the decision is made.

May a client ban AI use on their matter?

Yes. Engagement letters and procurement terms increasingly say something about AI, from a duty to notify to a ban on external AI services. Such a term overrides the firm's list: a tool that is allowed for the firm can be not allowed on that one matter. Check it per matter, and record it where the rest of the team on the file sees it.

Portrait of Rens Timmermans of BeeSensible

Would you rather someone showed you?

Rens, BeeSensible

I am happy to give you a twenty-minute tour. Leave your details and I will find a moment that suits you. Calling or emailing works just as well.

Rather talk to someone first? Book a call.

Want a live demo, or a quote for 100+ users? Leave your details and we'll be in touch within one business day.