Are our old contracts worth anything to AI companies, and can we even share them?

Contract archives can interest AI buyers, especially signed agreements kept with redlines and proposals. As of October 2026, nine tracked providers explicitly name these records. Each agreement’s confidentiality terms affect which files belong in a licensing package.

Why do AI companies ask for contracts?

The agreements as they were negotiated and kept, rather than clean templates. Appen’s systems-of-record category covers ledgers and contracts with the “negotiated exceptions of a live business”, lists “Contracts & e-signature” among its typical sources, and counts “Executed agreements and redlines” among what it extracts (as of October 2026). micro1 puts “contract workflows” and “review processes” under legal and contracts in its list of operational data that can contribute.

Contracts also explain the money. Troveo’s finance card groups them with the books: “Ledgers, invoices, and contracts — how the money actually moves”. A full contract file holds what was offered, what the other side changed, what both signed, and the invoices, change orders and renewals that followed.

What makes one contract file worth more than another?

Keeping the negotiation and what came after it, going by what providers name:

• Drafts and redlines kept with the signed copy. Appen pairs executed agreements with their redlines. How long drives keep earlier versions is on the documents page.
• Proposals with their outcome. Miro Advisory’s CRM card lists “proposals, wins, losses and customer outcomes” together, so keep proposals where they can be matched to the win or loss.
• Contracts tied to the work and the bills. A customer agreement whose invoices and change orders can be found beside it fits Troveo’s description above; the invoices and purchase orders page covers that side.

Blank templates, and forms bought from a publisher, carry none of that negotiation, and the publisher’s license governs them. The same signed PDF saved in the e-signature account, an email and a drive is one contract, not three.

Which buyers say they want contracts?

Nine of the 25 providers Data Licensing Report tracks, as of October 2026. A provider is listed only if its own pages name contracts, a kind of contract (agreements, proposals, leases, statements of work) or a contract system such as e-signature; all 25 were held to that test. An illustration used to explain de-identification does not count as an accepted category.

Provider	What it is	What its own pages name
Appen	Licenses directly	“Contracts & e-signature” among typical sources; “Executed agreements and redlines” among signals (source)
FileYield	Listing marketplace; buys nothing	Contracts in both its documents and legal categories (source)
License My Data	Introducer	Names contracts in its description of company operating history (source)
LH2 AI Labs	Licenses directly	“Filings · Contracts” in its finance and legal vertical (source)
Mercor	Licenses directly	Contracts under legal, leases under real estate, SOWs under professional services (source)
micro1	Licenses directly	Contract workflows and review processes under legal and contracts (source)
Miro Advisory	Introducer; does not buy data	Proposals, on its CRM and sales card (source)
Polyshares	Licenses directly	Contracts among SharePoint’s “controlled documents”, and proposals among what Drive holds (guides checked September 2026)
Troveo	Licensing agent	Contracts on its finance and operations card (source)

General documents, shared-drive access and a legal-industry label alone do not qualify a provider for this table. The documents page covers broader document programs; the buyers directory explains how each channel pays.

Where are our contracts, and are the old ones still there?

In several places at once, each keeping its own copy:

• E-signature accounts. A DocuSign employee posting as a community moderator said on April 13, 2022 that retention depended on administrator-enabled Document Retention or Targeted Purge. That historical forum answer described a 14-day purge queue. On January 2, 2026, another DocuSign employee recommended Retrieve as a paid add-on for bulk downloads. These are dated community answers; check your account’s retention settings and Retrieve entitlement before planning an export.
• Drives and email. Signed PDFs and Word drafts sit in shared drives, whose version and trash limits are on the documents page, and as attachments in mailboxes (email).
• Business software. Proposals and quotes live in the CRM (CRM records), change orders in construction project software (construction), and leases in property-management systems (property management).

Contracts signed before e-signature may survive only on paper or as scans, so note which years exist in which form.

Our contracts say their terms are confidential. Can we share them at all?

Only as far as each clause allows, and wording varies by contract. Common Paper’s standard agreements, published free to use under CC BY 4.0, show the parts such clauses are built from:

• What counts. Its Mutual NDA (Version 1.0) covers information the discloser identifies as confidential or that “should be reasonably understood as confidential or proprietary due to its nature and the circumstances of its disclosure”.
• The deal itself. Its Cloud Service Agreement (Version 2.1) adds “Confidential Information includes the existence of this Agreement and the information on each Cover Page”. The Order Form is a Cover Page, and the agreement defines Fees as “the applicable amounts described in an Order Form”.
• Permitted use. The NDA has the recipient “use Confidential Information solely for the Purpose” and disclose it to third parties only with written approval, apart from its own representatives who need to know.
• What falls outside. Information that is public, already known, received from a third party without restriction, or developed independently.
• How long. Its editable Cover Page offers a fixed confidentiality term, with a one-year placeholder measured from either the effective date or last disclosure, and separate treatment for trade secrets. An indefinite term is another option. The completed agreement sets the period.
• After the relationship ends. The Cloud Service Agreement has each side return or destroy the other’s confidential information, and for copies kept under backup or record-retention policies, its privacy and confidentiality sections “will continue to apply to retained Confidential Information”.

The relevant wording is in the agreement your company actually signed. Is it legal to sell company emails and records? lists the clauses that affect scope; have any unresolved confidentiality restriction interpreted before including that file.

Providers plan for restricted contracts. Troveo’s September 15, 2026 guide for closing startups says “anything a customer contract restricts” comes out before delivery; Corpus asks sellers to keep material under legal hold or NDA restriction out of scope; Avelence’s profile form has owners flag “Customer confidentiality” among known restrictions; and Miro Advisory’s FAQ says de-identification does not resolve a contractual restriction on the underlying records.

Are the emails with our lawyer about a contract privileged?

Some may be, and they are a separate pile from the contracts. Cornell’s Legal Information Institute explains that attorney-client privilege “protects confidential communications between a lawyer and their client that relate to the client’s seeking of legal advice or services”, and that a third party present during the communication can compromise that confidentiality unless the third party is essential, such as an interpreter. The signed contract went to the other side; the advice you received about it did not.

In a contract file, look for legal advice in:

• emails to and from your lawyer about terms, renewals or a dispute;
• drafts carrying counsel’s comments that were sent only to you;
• memos on risk written before signing.

Telegraph Lab, a provider affiliated with Data Licensing Report, says privileged and third-party information needs additional review and may be excluded, and License My Data lets owners exclude legal records along with board, finance, HR and M&A. The lawyer who advised on each deal can say which files are privileged; leaving out records shows how to describe them, and is it legal? covers waiver.

Whose personal information is in a contract file?

The signers’, and anyone the contract is about:

• Signers and notice contacts. Names, titles, emails and addresses.
• E-signature records. The University of Texas at Austin’s DocuSign guide says the Certificate of Completion shows each signer’s email address and “The IP address of the signer when signing occurred”.
• Employees. Employment agreements, non-competes and employee NDAs are HR files, which Corpus asks sellers to leave out of scope.
• People as counterparties. Residential leases and agreements with sole proprietors name individuals; property management covers tenant files.

Replay illustrates de-identification with an image caption naming “A PDF agreement”; that illustration does not establish a contract-buying program, and Telegraph Lab’s process finds “Names, contact details, account IDs, and confidential information” and replaces identifiers “with consistent placeholders”. Replacing names leaves prices and terms in the text, which is why the clause question comes first; what “anonymized” means covers the rest. The sample tool does not read PDF or Word files, so contract text is checked by the provider or your own reviewer before anything leaves.

What could our contracts be worth?

The sources in the buyer table do not establish a price for your contracts alone. The ranges providers publish for whole companies are on How much is my company’s data worth?. To describe a contract set to a buyer, count signed agreements by kind (customer, vendor, lease, employment) and by year, how many keep their drafts or redlines, and how many carry a confidentiality clause your lawyer has not yet cleared.