# Preparing legal documents for an AI data license > A legal archive only becomes a license once it is clean on three fronts – client consent, material owned by others and personal data. Here is how firms generally sort matters, why discovery and sealed material usually stay out, and what is removed from documents before anything is shared. Before a legal archive can be licensed, it has to be sorted. Not every closed matter can be included, and the ones that can still have to be cleaned. This guide covers the three screens most legal archives go through, the material that usually stays out, how documents are anonymized and a checklist to see where your firm stands. *This guide describes how things generally work today. It is not legal advice. Every archive and every deal is different – what applies depends on your matters, your clients and the orders in each case, and is assessed case by case. And that's what we're here for – we do the groundwork with you, involve the right experts when needed and put good deals on the table.* ## Why preparation decides whether a license happens A buyer has to be able to rely on the material it trains on. For legal data, that means three things have to be clear before delivery: 1. **Consent and confidentiality.** Either the material holds no client information, or the client has given informed consent. 2. **Ownership by others.** Nothing in the selection belongs to another party or is restricted by a court order. 3. **Personal data.** Names and other personal information have been removed or replaced. An archive that passes all three is easier to license and usually worth more. How ownership and confidentiality generally work is covered in [can law firms license client files for AI](https://corpusmarket.com/guides/can-law-firms-license-client-files-for-ai.md). For how buyers price legal data, see [how much is law firm data worth](https://corpusmarket.com/guides/how-much-is-law-firm-data-worth.md). ## Sorting matters by consent A practical first step is to sort the archive into groups: - **Firm-owned material** – templates, forms, clause libraries, practice notes and training material with no client information. Usually the first part of a license. - **Matters where consent can be asked for** – closed matters for clients the firm still has a relationship with, where the engagement terms don't rule it out. - **Matters that stay out** – clients who can't be reached or are unlikely to agree, engagement terms or outside counsel guidelines that restrict use, and matters that are too sensitive. How a consent request is worded, and which clients it goes to, is decided case by case. Many firms start with a small number of matters and expand from there. ## Discovery productions and protective orders Litigation files often hold a lot of material the firm didn't create: - **Documents produced by other parties** in discovery, including custodial email and chat from the other side. - **Expert reports, deposition transcripts and exhibits** that may be designated confidential. - **Data room contents** from transactions, provided under confidentiality agreements. Protective orders under Federal Rule of Civil Procedure 26(c), and their state equivalents, often limit the use of designated material to the litigation itself and require it to be returned or destroyed when the case ends. Even without an order, documents produced by another party aren't the firm's to license. In practice, this material is left out of a license, however much of it there is in the review platform. ## Sealed records and sensitive matters Some matters carry restrictions beyond the parties' agreements: - **Sealed filings and sealed cases**, where a court has limited access to the record. - **Juvenile, family and mental health matters**, where statutes often restrict disclosure. - **Grand jury and government investigation material**, which carries its own secrecy rules. - **Matters under government contracts or security requirements.** These are flagged early and normally excluded. ## Personal data in matter files Matter files are full of personal data – not only about clients, but about opposing parties, witnesses, employees and family members. Some of it is sensitive: medical records in personal injury and employment cases, financial records in family and commercial matters, Social Security numbers in nearly everything. Where health records came to the firm from a healthcare client, a business associate agreement under HIPAA may also apply. No deal involving health data moves forward without legal review. More on how personal data is handled in [data privacy and AI training](https://corpusmarket.com/data-privacy-and-ai-training.md) and in our guide on [anonymized vs de-identified data](https://corpusmarket.com/guides/anonymized-vs-de-identified-data.md). ## How to anonymize legal documents Once a selection is agreed, the documents are cleaned before delivery. A typical clean-up covers: - **Names.** Parties, witnesses, judges where relevant, counsel and firm staff are replaced with consistent placeholders, so "Party A" stays "Party A" across the whole matter. - **Identifiers.** Case and docket numbers, court names where they identify the matter, addresses, dates of birth, account and policy numbers, and Social Security numbers. - **Signatures and letterheads**, including the firm's own if the buyer doesn't need them. - **Document metadata.** Author fields, tracked-change authors, comment authors, file paths and document management IDs. - **Embedded content.** Images, scanned exhibits and attachments, which are checked or removed. Consistent placeholders matter. A model learns little from a document where every name is just "[REDACTED]", but a lot from one where the parties stay distinct. We work with our partners and the buyer to clean documents in the best possible way to close the deal. ## Seller checklist You don't need to answer all of these before an assessment, but the more you know, the faster it goes: 1. Which practice areas and years does the archive cover? 2. Roughly how many closed matters, and how many documents per matter? 3. Which systems hold them – document management, email, review platforms, shared drives? 4. Are drafts and version history kept? 5. Is there firm-owned material such as templates, clause libraries and practice notes? 6. Do your engagement letters or outside counsel guidelines say anything about using matter documents? 7. Which clients might be asked for consent? 8. How much of the archive is discovery from other parties or data rooms? 9. Are there sealed matters, or matters under protective orders? 10. Which practice areas hold a lot of health, financial or other sensitive personal data? ## What we handle and what your counsel handles We do the groundwork: mapping the archive, sorting matters into likely groups, finding buyers, negotiating terms and coordinating clean-up and delivery. Questions about your own professional obligations – your state's version of Rule 1.6, specific engagement terms or orders in particular cases – are ones many firms take to their ethics counsel or general counsel, and we bring in the right experts when a question calls for it. The final call on what is included is always yours. Throughout, you stay in control. Nothing is shared without your approval, and you approve every deal. Licenses are non-exclusive by default and limited in scope, time and use, so the firm keeps ownership of its documents. Read more about how to [license your firm's legal archive](https://corpusmarket.com/sell-law-firm-data.md). ## Summary A legal archive is sorted on three fronts before it can be licensed: client consent and confidentiality, material owned by others and personal data. Firm-owned material usually comes first, client matters are included only with consent, and discovery productions, data rooms, sealed records and protected matters usually stay out. Selected documents are cleaned with consistent placeholders and stripped of identifiers and metadata. You don't need to clean anything before an assessment, and every firm and every deal is assessed on its own. ## Frequently asked questions ### Can documents produced in discovery be licensed? Usually not. Documents a firm receives in discovery generally belong to the producing party, and stipulated protective orders often limit their use to the case and require return or destruction when it ends. They are normally left out of a license, even when they still sit in the firm's review platform. ### How are legal documents anonymized for AI training? Party, witness and counsel names are replaced with consistent placeholders, and case numbers, courts, dates, addresses, account numbers and signatures are removed or generalized. Document metadata, tracked-change authors and comments with names are stripped too. The aim is that a document can't be traced back to a client or a matter. ### What about matters with health or financial data? They need far more care. Matters can hold medical records, financial statements and other sensitive personal data about clients, opposing parties and witnesses. Some practice areas, such as personal injury or family law, are often left out for that reason, and no deal involving health data moves forward without legal review. ### Do we need to clean up the archive before an assessment? No. It is better not to. Keep documents, versions and folder structures as they are. The assessment works from a description of the archive, and the selection and clean-up happen only after you have approved a deal.