Law firm data and AI training: what a firm can license

By Cody Houser, Founder · Last updated October 24, 2026

The short answer: A law firm can license material it owns that contains no client information, such as its own templates, practice guides, training materials, and de-identified operational records. Client documents and matter email are different: ABA Model Rule 1.6 bars revealing any information relating to a representation without informed consent, and ABA Formal Opinion 512 says boilerplate engagement letter language is not enough for AI use.

This guide is for managing partners and general counsel who have been asked, or are wondering, whether the firm’s archive has value to AI developers. It walks through the rules that decide what can leave the firm, sorts common material by risk, and ends with a sequence you can follow with your ethics counsel. It explains the rules; it is not legal advice for your jurisdiction.

Why AI companies want legal data

Legal AI is built on documents. When vLex and Fastcase announced their merger in 2023, Fastcase’s CEO called the combined library of more than 1 billion legal documents the “ultimate training data set for legal AI”. That library is public law: opinions, statutes, regulations, dockets, and filings. In August 2023, 273 Ventures launched the Kelvin Legal DataPack, which it described as “the first large-scale legal dataset with clear provenance and commercial use rights”, over 150 billion tokens drawn from sources such as EDGAR and PACER.

Both examples show the same thing: public legal text is already being packaged at scale. What a firm holds that the public record doesn’t is how the work gets done: the drafts, the markups, the checklists, the reasoning between versions. That is also exactly where client information lives, which is why the firm’s first question is what it may share, not what a buyer wants. For the wider market, see how companies sell data to AI developers.

Rule 1.6: the scope of “information relating to the representation”

Model Rule 1.6(a) reads: “A lawyer shall not reveal information relating to the representation of a client unless the client gives informed consent, the disclosure is impliedly authorized in order to carry out the representation or the disclosure is permitted by paragraph (b).” None of the paragraph (b) exceptions, which cover things like preventing death or fraud and complying with a court order, covers licensing data for profit.

Three comments to the rule decide most licensing questions:

It is not limited to secrets. Comment 3 says the rule applies “not only to matters communicated in confidence by the client but also to all information relating to the representation, whatever its source.” Research memos, opposing counsel’s letters, and deal terms all count.
Indirect disclosure counts. Comment 4 extends the rule to disclosures that “could reasonably lead to the discovery of such information by a third person.” Hypotheticals are allowed only where there is “no reasonable likelihood” the listener can identify the client or situation. That is the bar a de-identified dataset has to clear.
It survives the engagement. Comment 20: “The duty of confidentiality continues after the client-lawyer relationship has terminated.”

Rule 1.6(c) adds a security duty: reasonable efforts to prevent “unauthorized disclosure of, or unauthorized access to” client information. Handing an archive to an outside party for review or de-identification engages that duty before any license is signed.

ABA Formal Opinion 512 and informed consent

Formal Opinion 512, issued July 29, 2024, is the ABA’s guidance on lawyers’ use of generative AI. It concludes that because many self-learning AI tools can surface one client’s information to other users, “a client’s informed consent is required prior to inputting information relating to the representation into such a GAI tool.”

The opinion is about using AI tools in a representation, not selling training data. But licensing client material to an AI developer goes further than using a tool: the stated purpose is to let the developer learn from it. The opinion’s consent standard is the floor to plan around. For consent to be informed, the client needs “particulars about the kinds of client information that will be disclosed, the ways in which others might use the information against the client’s interests,” and the extent of the risk that later users of the tool will have access to it. And: “merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient.”

Rule 1.0(e) defines informed consent as agreement “after the lawyer has communicated adequate information and explanation about the material risks of and reasonably available alternatives.” In practice, that means a specific, matter-level conversation with each client, not a clause added at intake.

Former clients and Rule 1.9(c)

Most of a mature firm’s archive belongs to former clients. Rule 1.9(c) says a lawyer “whose present or former firm has formerly represented a client in a matter” shall not “reveal information relating to the representation except as these Rules would permit or require with respect to a client,” and shall not use it to the former client’s disadvantage unless it has become generally known.

So closed matters need the same consent as open ones, and getting it is harder: clients have been acquired, dissolved, or moved to other counsel, and contacts have changed. Many firms will find that consent is practical for a handful of long-standing clients and impractical for the rest, which shapes what is worth offering.

Who owns the client file

Ownership and confidentiality are separate questions. Even where the firm owns a document, Rule 1.6 still controls what it may reveal. Ownership matters for a different reason: it decides what the firm must hand back and what it may retain.

Rule 1.16(d) requires a lawyer to surrender “papers and property to which the client is entitled” at the end of a representation, but the Model Rules don’t define that phrase. ABA Formal Opinion 471 (2015) noted that most jurisdictions apply an “entire file” approach, presuming the lawyer returns everything, while a minority use an “end product” approach. Idaho codified the entire-file approach in 2025, joining Washington, Oregon, and Alaska bar opinions. Separately, Rule 1.15(a) requires lawyers to keep client property “separate from the lawyer’s own property” and to safeguard it. Check your state’s file rule before assuming any client document is the firm’s to license.

Privilege, work product, and confidentiality

Comment 3 to Rule 1.6 describes three related bodies of law: the attorney-client privilege, the work product doctrine, and the ethical duty of confidentiality. Privilege and work product apply when evidence is sought through compulsion of law. Federal Rule of Civil Procedure 26(b)(3), for example, ordinarily shields documents “prepared in anticipation of litigation or for trial” from discovery. Confidentiality is broader and applies everywhere else.

Two consequences for licensing. First, material that isn’t privileged can still be confidential, so “not privileged” doesn’t mean “licensable.” Second, as Frankfurt Kurnit notes in its guidance on selling company data for AI training, a sale may waive privilege over legal communications. Exclude privileged communications entirely.

Protective orders, sealed filings, and public records

Litigation files often contain the other side’s documents, produced under a protective order. Courts can order under FRCP 26(c)(1)(G) that a trade secret or other confidential commercial information “not be revealed or be revealed only in a specified way.” Those orders typically limit use to the litigation, and sealed filings carry the court’s own restrictions. Neither client consent nor firm ownership overrides a court order.

Public filings are not automatically free either. ABA Formal Opinion 480 (2018) says the duty of confidentiality applies even when information about a representation is found in a court record or other public record. The opinion addressed blogging, but its reasoning reaches any disclosure of a firm’s own matters.

Firm know-how, departing lawyers, and other firm material

The cleanest licensing candidate is material the firm created for itself: form banks and templates built from scratch, practice guides, training decks, checklists, internal procedures, and operational records such as staffing and workflow data with client details removed. Watch for templates that began life as a client’s deal documents; those may carry client terms or another firm’s drafting.

Lawyer departures add two questions. Rule 1.9(c) binds lawyers whose former firm represented a client, so a partner who has left is still bound, and a lateral who joined brings duties to the old firm’s clients. And clients choose their counsel: ABA Formal Opinion 99-414 says both the departing lawyer and the firm must make sure clients are told of a departure so they can decide whether the work stays or moves. A client whose matters have left the firm is a poor candidate for consent. Partnership agreements may also say who owns know-how a departing partner created, so read them before including that material.

What a law firm can license: material by material

Material	Can it be licensed?	Conditions
Firm-created templates and practice guides	Often yes	Built in-house with no client terms; check partnership agreements on authorship
Internal operations records (staffing, workflow, the firm’s own finances)	Often yes	Remove client names, matter details, and personnel data; billing narratives are client information
Client documents and work product	Only with informed consent	Specific, informed consent per Rule 1.6 and Opinion 512; former clients under Rule 1.9(c); state file rules
Matter email and chat	Rarely	Dense with client information, third-party data, and privileged content; consent from every client involved
Public filings from the firm’s matters	With consent	Opinion 480: public-record status doesn’t end confidentiality; check sealing orders
Privileged communications	No	Disclosure may waive privilege; exclude
Material under protective order or seal	No	Court order controls; client consent can’t override it

For what makes the licensable categories valuable to buyers, see the data value checklist and how much company data is worth.

Steps for a managing partner

1. Bring in ethics counsel first. Before anyone describes the archive to a buyer in detail, have your firm’s general counsel or outside ethics counsel confirm your state’s versions of Rules 1.6, 1.9, and 1.16 and any local AI opinions.

2. Inventory by system. Document management, email, chat, billing, matter management, and knowledge management. For each, note date range, volume, and whether it holds client information.

3. Separate firm material from client material. Tag knowledge-management content that was built in-house versus adapted from client matters. This is usually the slowest step and the most valuable.

4. Flag hard exclusions. Privileged communications, documents under protective order or seal, matters with confidentiality agreements, and HR and personnel records.

5. Decide whether consent is realistic. If you want to include any client material, list the clients, draft a matter-specific disclosure that meets Opinion 512’s standard, and accept that some will say no.

6. Describe, don’t send. The first conversation with a buyer or broker should cover practice areas, firm size, years of history, and systems. No documents, credentials, or client information.

7. Control the agreement and transfer. Scope, permitted uses, de-identification standards, security, deletion, and audit rights. Use the AI data licensing agreement checklist. If the firm is dissolving, see licensing data when a business is closing.

Questions people ask

Can law firms use client data for AI training?

Only with the client’s informed consent, in most cases. ABA Model Rule 1.6 bars revealing any information relating to a representation without consent, and ABA Formal Opinion 512 says informed consent is required before client information goes into a self-learning AI tool. Boilerplate in an engagement letter is not enough.

Can a law firm sell its data?

A firm can license material it owns that contains no client information, such as firm-written templates, practice guides, training materials, and operational records with client details removed. Client material needs informed consent, and some material, such as documents under a protective order, may not be shareable at all.

Does removing client names make documents safe to license?

Not by itself. Comment 4 to Rule 1.6 says the duty covers disclosures that could reasonably lead a third person to discover protected information. Facts, deal terms, dates, and context can identify a client or matter after names are gone.

Is information in a public court filing still confidential?

Under the ABA’s view, yes. ABA Formal Opinion 480 says the duty applies even to information found in a court record or other public record. Sealed filings and documents produced under a protective order carry their own court-imposed limits.

Is attorney-client privilege the same as confidentiality?

No. Privilege and work product are evidence rules that apply when a court or opponent seeks information. The ethical duty of confidentiality under Rule 1.6 is broader and covers all information relating to the representation, whatever its source.

Does the duty to former clients end when the matter closes?

No. Comment 20 to Rule 1.6 says confidentiality continues after the client-lawyer relationship ends, and Rule 1.9(c) bars revealing a former client’s information, or using it to the former client’s disadvantage, except as the rules permit.

How Hyperdrive Data fits

Hyperdrive Data is a broker, and legal is one of the industries current buyers are interested in. A first conversation is a description of the firm and its systems, with no files or client information. If there’s a fit with current buyer requirements, the buyer’s proposed agreement sets the terms, and you and your counsel decide whether to proceed. See whether your firm fits.

Sources

This guide is general information, not legal, tax, or financial advice. Review any proposed agreement with your own advisers.

Related guides

Check whether your company is a fit. Start with a description of your records; no files needed.