Can a Law Firm License Its Data to AI Companies?

Can a law firm license its data to AI companies?

A law firm can license material it created that carries no information relating to a client's representation, and should treat matter files as unavailable without each affected client's informed consent. That is the working position, and it rests on an inference rather than an opinion on point, which is worth stating plainly because nobody else does.

No ABA or state ethics opinion addresses a law firm licensing client files to an AI company. The nearest authority is ABA Formal Opinion 512, issued on July 29, 2024, which covers lawyers' own use of generative AI. As Frankfurt Kurnit Klein & Selz summarized it in January 2025, the opinion concluded that "a client's informed consent is required before inputting their confidential information into a self-learning GAI tool", and that informed consent "requires the lawyer's explanation of the risk, not merely boilerplate provisions in an engagement letter". We cite the summary rather than the opinion because the ABA's own site refused access when this page was checked.

If a lawyer needs informed consent to put client information into a tool the firm controls, a transfer of the same information to a third party for training is not a lesser disclosure. The confidentiality duty under Model Rule 1.6 is wider than most data buyers' questions assume. The comment to the rule, in Minnesota's adoption of it, says the rule "applies not only to matters communicated in confidence by the client" but "to all information relating to the representation, whatever its source."

So the useful question is not whether a firm may do this. It is which part of a firm's archive contains no information relating to any representation. That part exists, it is the part that makes one firm's work different from another's, and the rest of this page is about identifying it and what it is worth.

Why are AI companies bidding for Weil Gotshal's data, and does that apply to your firm?

AI companies are bidding for a dissolving firm's assets, which is a different transaction from a going concern licensing its files, and the distinction is being lost in the coverage. Weil Gotshal & Manges is winding down, and its data is being treated as an asset in that wind-down.

The American Lawyer reported on November 4, 2026, in a piece by Amanda O'Brien, that "at least two AI-native law firms, Scissero and Vetta, both based in London, have put in bids for Weil Gotshal & Manges' data". The same report notes that "Client confidentiality requirements, however, could pose an obstacle to any firm attempting to sell their internal data". Law360 Pulse reported the same day, in a piece by Jake Maher, that as Weil "approaches dissolution" AI-focused firms are "making bids to scoop up some of the firm's data assets and intellectual property". Both reports are paywalled after the opening, and no dollar figure appears in the visible text of either.

Three things follow for a firm that is not dissolving. The assets in play are reported as brand, trademark and internal data rather than client matter files, and the confidentiality obstacle the reporting flags is the same one on this page. A dissolution changes the commercial calculus, because there is no future practice whose optionality is being given up, but it does not change the ethical duty, which survives the firm. And the valuation question is open: the reporting describes bids rather than a settled price.

The headlines are evidence that buyers want this material. They are not evidence that a live firm can sell matter files.

What does ABA Formal Opinion 512 mean for a data license?

ABA Formal Opinion 512 means that client information cannot move into a training context without informed consent that explains the actual risk, and that an engagement letter clause will not do the work. The opinion is about a lawyer's own use of generative AI, so applying it to a data license is an extension, and it is the extension every careful reading of it supports.

The opinion was issued on July 29, 2024 by the ABA Standing Committee on Ethics and Professional Responsibility. Its reasoning turns on self-learning tools: information entered into one can surface in another user's output, which is why consent is required even when the tool is used only inside the firm. A license that delivers records to a lab for training is that concern in a more direct form, since the material is handed over deliberately and permanently.

Two features of informed consent make the client-by-client route impractical for most firms. It has to explain the extent of the risk that later users of the tool will have access to information relating to the representation, which means explaining a technical process whose outcome nobody can bound. And it has to be obtained from each affected client, including former clients whose matters are closed, because as the comment to Rule 1.6 puts it, "The duty of confidentiality continues after the client-lawyer relationship has terminated."

One more point belongs here because buyers raise it. De-identification does not answer the question. Information relating to a representation can be identifiable from the facts of a matter alone, and the duty attaches to the information rather than to the name on it. A buyer offering to anonymize is describing its pipeline, not resolving your obligation.

What can a law firm actually license?

A law firm can license the material that makes the firm good at its work, as long as no part of it relates to a client's representation. In a practice built on precedent, that is a substantial archive and it is the one buyers in this market have paid for elsewhere.

Four categories usually survive the test:

• Template and model documents the firm drafted, including clause libraries and fallback positions.
• Playbooks and negotiation guidance, meaning the firm's own written position on how a point is taken and traded.
• Internal training material, know-how memoranda and practice notes written as general guidance rather than advice on a matter.
• The firm's own operational records: how work was staffed, reviewed, escalated and resolved internally, with matters and clients removed rather than masked.

That last category is close to what the general buyers describe. micro1 lists SOPs, process documentation, internal knowledge bases, workflows, templates, CRM records and support tickets on its data partnerships page, read on November 10, 2026, names legal and contracts among its industries, and adds a category it calls agent trajectories, the record of how a task was actually carried out. Handshake lists Legal among its industries on its data partnerships page, read the same day, and says "Your documentation, workflows, and project history capture how your business operates".

Two cautions on the surviving categories. A model document drafted for one deal and never generalized can still carry that deal's fingerprints, so templates need reviewing rather than exporting. And a playbook built from one client's recurring instructions is that client's information in a general wrapper. The test is whether the material would read the same if the firm had never acted for any particular client.

What has to stay out of a law firm's data license?

Everything relating to a client's representation stays out, and that is wider than privileged communications. Model Rule 1.6 reaches all information relating to the representation whatever its source, which is the standard to apply when scoping anything.

These stay out of any scope you describe to a buyer:

• Matter files in full, including pleadings, correspondence, memoranda and advice.
• Discovery material, which frequently belongs to a third party and is commonly covered by a protective order that no firm can override by agreement with a buyer.
• Deal documents from live or closed matters.
• Client lists, engagement terms and billing records.
• Anything subject to a confidentiality undertaking given to another party.
• Work product created for a specific client.
• Conflicts data, which is a map of who the firm acts for.

Also out on the standard grounds that apply across this lane: HR and payroll records, health information, and call or video recordings made without all-party consent.

Ask any buyer in writing whether its proposed scope reaches matter content in any form, including redacted, aggregated or synthetically rewritten. The answer tells you whether the conversation is about your know-how or about something you cannot sell. What remains is the firm's own product, and the discipline of separating it first is what lets a managing partner have this conversation without a risk committee meeting after every call.

What is a law firm's know-how worth to AI companies?

No buyer publishes a figure for law firms, and the published numbers nearby are ceilings rather than quotes. micro1 shows $100K to $500K for qualified partnerships, $500K to $3M for large-scale or multi-team data and $3M+ for highly unique data. Scale AI shows "$10K to $1M+" on its data partnership page, labelled illustrative and scaling with cadence. Handshake leads with "Earn $100K to $4M+". Mercor publishes no range and says compensation is based on data volume, the tools you connect and the depth of your data. All five pages were read on November 10, 2026.

The first row contains the practical problem. The buyer most explicit about wanting legal data wants case files and discovery documents, which is the category a firm cannot supply. A firm offering templates, playbooks and internal workflow instead is offering something real and something smaller, and the number should be expected to reflect that rather than the $3M ceiling in micro1's published tiers. Every buyer here is read side by side in which AI company pays the most for business data.

Terms matter as much as price here. Polyshares describes its published licenses as perpetual and exclusive, and a model cannot unlearn what it has been trained on, so a perpetual grant over a firm's precedent library is a permanent decision about the firm's own competitive position. Our page on exclusive versus non-exclusive licenses covers how to limit it by term, field of use or scope, and is it legal to sell your company data to AI companies covers the general privacy law underneath all of it.

Frequently asked questions

Related resources

• Can an accounting firm license its data to AI companies?
• Can I sell my company's data to AI companies?
• Is it legal to sell your company's data to AI companies?
• How BLACKSIG helps companies license their data