Are our shared drives and documents worth anything to AI companies?

Often, yes, for the procedures, proposals, reports and templates your company wrote, more so with versions and authors intact. Clients’ files and documents received under an NDA need sorting out first. As of October 2026, 23 of the 25 providers Data Licensing Report tracks name documents or a drive.

What do AI companies use company documents for?

To train and test models on how your company says its work should be done, and on the work it produced. Appen’s first category of records is “Procedures, knowledge bases and templates that define how a task is meant to run, including the exceptions nobody ever wrote into the official version”, drawn from docs, wikis, cloud storage and training material. micro1 asks for “standard operating procedures (SOPs), process documentation, internal knowledge bases, workflows, templates, operational frameworks” (both as of October 2026).

Finished work counts too. Polyshares’ Google Drive guide says “Drive holds the documents a company writes to run itself: proposals, procedures, project plans, pricing sheets and reports”, its SharePoint guide that “SharePoint sites are where teams keep controlled documents: policies, procedures, project files, contracts and quality records” (both checked September 2026), and Mercor’s examples for professional services are client decks, SOWs and project files.

Which documents are worth more, and does version history matter?

Yes. Troveo, Appen and Avelence name versions or revisions among what they want, and Nyne and Scale AI ask for authorship and timestamps, as of October 2026.

Raises the value	Where a provider says so
Versions and authors kept with the file	Troveo: “Contents plus the metadata: authorship, versions, timestamps”
The reasons behind changes	Appen lists “Revision history and rationale” among the signals it takes from documentation
Drafts as well as finals	Avelence: “Working documents, revisions, and shared files that show how knowledge develops over time”
Documents tied to the rest of the record	Appen: “Connected systems are worth considerably more than isolated exports”

Each task in idler’s CorpLaw benchmark, drawn from a real law firm’s anonymized record, includes “document versions, negotiation email threads, defined terms, and client instructions”, and Telegraph Lab’s deal card for an engineering firm lists “revisions” beside the drawings (below).

Less useful are duplicates (one file saved in five folders is still one document), manuals and standards your company downloaded but did not write, and copies that lost their authors and dates on export, the metadata Troveo, Nyne and Scale AI list.

Which buyers say they want documents?

All but two of the 25 providers Data Licensing Report tracks, as of October 2026. Each one below names documents, a document type such as SOPs or specs, or a file system such as Google Drive, SharePoint, Dropbox or Box on its own pages; Defined.ai and the Google Content Offer Pilot name none of these.

Provider	What it is	What its own pages name
Appen	Licenses directly	Docs and wikis, cloud storage, intranet (source)
Avelence	Introducer; does not buy data	Drive, OneDrive, Dropbox (source)
Corpus	Licenses directly	“documents”; Dropbox, SharePoint, Notion (source)
FileYield	Listing marketplace; buys nothing	Contracts and inspection reports in its documents category
Handshake AI	Licenses directly	Google Drive, Dropbox, Notion (source)
DataVendor (HUD)	Listing marketplace; buys nothing	Zip bundles of documents; internal-doc snapshots from closing startups
idler	Licenses directly	Document versions from a law firm’s record
LH2 AI Labs	Licenses directly	“workflows · documents” (source)
License My Data	Introducer	“SOPs, playbooks and internal documentation”
Mercor	Licenses directly	Drive, Confluence, Notion (source)
micro1	Licenses directly	SOPs, process documentation, templates (source)
Miro Advisory	Introducer; does not buy data	Google Drive, SharePoint, Dropbox (source)
Nyne	Licenses directly	Word, Docs, Slides, Excel, Sheets, Drive, SharePoint (source)
Polyshares	Licenses directly	Drive and design files in its list; export guides for SharePoint, OneDrive, Dropbox and Box
Replay	Licenses directly	“Internal docs” on its home page; Google Docs, Sheets and Notion among sources
Scale AI	Licenses directly	Word, Docs, Slides, Excel, Sheets, Drive, SharePoint (source)
Scrimdata	Licenses directly	“document trees”
Sell My Business Data	Introducer	Drive, SharePoint, Dropbox, Box (source)
SimpleClosure (AssetHub)	Wind-down platform; said on September 3, 2026 it also takes operating companies	Documents; Google Drive, Notion (source)
Sunset	Wind-down platform; buys from startups that are closing	“product specs”
Telegraph Lab (affiliated)	Licenses directly	Google Drive, Dropbox, Box: PDF, DOCX, XLSX, CAD (source)
Troveo	Licensing agent	Google Drive, OneDrive, SharePoint, Dropbox (source)
Turing (Project Lazarus)	Licenses directly, from startups that shut down	“design docs” and internal wikis

The buyers directory compares how each one pays.

Where do our files live, and how much history do they keep?

In the drive, until someone deletes them; earlier versions and deleted files are kept for shorter periods that differ by system and plan. As of October 2026:

System	Earlier versions	Deleted files
Google Drive	Docs, Sheets and Slides keep their own history; for other files, “A version might be permanently deleted after 30 days or if there are 100 newer versions” unless marked Keep forever	“Files you move to the Trash are deleted forever after 30 days”
SharePoint and OneDrive	Up to the limits admins or site owners set; new organization limits apply only to new libraries	Recycle bins keep items 93 days
Dropbox	30 days on Basic, Plus and Family; 180 on Professional, Essentials, Business and Standard; 365 on Business Plus, Advanced and Enterprise (Dropbox)	The same plan periods
Box	25 versions on Starter, 50 on Business and Business Plus, 100 on Enterprise, unlimited on Enterprise Plus and Enterprise Advanced (Box)	Trash keeps items 30 days by default, and admins can change it

Accounts of people who have left are a separate gap. Google says “When someone leaves and an admin deletes their account, files they added or created in shared drives remain”, while files in a deleted user’s own My Drive go with the account unless transferred; a departed person’s OneDrive follows Microsoft’s unlicensed-account schedule; the Google Workspace and Microsoft 365 pages give the timings. Files from before the cloud may still be on an old server (old systems).

Exports treat versions differently. Microsoft Purview eDiscovery lets you choose “Latest version only”, the recent 10 or 100, or all versions of SharePoint and OneDrive files. Google Vault searches the latest version of each file; for Docs, Sheets and Slides you can add a version date, and then “you preview and export the contents of the last version saved before 12:00 AM on the specified date”. Polyshares asks for drive size in gigabytes and says version history does not count toward that figure, so note versions separately in your data inventory.

Some files in our drive are clients’ or other companies’. Can they go out?

Not until someone has checked who holds the rights. Four kinds of file in a company drive are not yours alone:

• Records you hold for clients. An accounting firm’s client workpapers, a law firm’s matter files, an agency’s client assets. Their rules start with the client relationship; see the accounting, law firm and marketing agency pages.
• Documents received under an NDA or contract. Drawings, specifications and price lists from customers, contractors and suppliers. Their terms govern use wherever the file sits, and Corpus’s out-of-scope list includes material under “active legal hold or NDA restriction”. Is it legal? lists what to read in each clause.
• Files owned outside the company. A Google Vault search of a person’s account includes “Files directly shared with the account, including files owned by users outside your organization”, so an export built person by person can carry other companies’ files.
• Material you bought or downloaded. Purchased templates, industry standards and manufacturers’ manuals come with their sellers’ licenses. Avelence makes the same point about software: “The rights in third-party and open-source material need separate attention”.

Where a drive is organized by client or project, these can be scoped out by folder (leaving out records), and the sample tool recounts without any top-level folder you untick.

What else in a shared drive needs keeping out?

Files about people, and what earlier versions still hold:

• HR and pay. Corpus asks sellers to leave HR, pay and performance files out of scope, along with “Government-issued identification numbers”, which turn up in scanned IDs, tax forms and payroll spreadsheets.
• Spreadsheets of contacts. A customer list saved as a sheet is CRM data by another name; the CRM page covers the privacy questions.
• What the final version removed. Comments, tracked changes and earlier versions can still hold a client’s name or a first price that the final copy dropped, so a review has to cover every version you deliver.
• Who could see what. Nyne and Scale AI ask for files with “access permissions”, which name the people and outside accounts a file was shared with.

Your lawyer can say which of these reach your drive. For what providers say they replace inside documents, see what “anonymized” means.

What could our documents be worth?

No provider publishes a price per document or per gigabyte, as of October 2026. Telegraph Lab (affiliated with Data Licensing Report) shows two past deals made up mostly of documents on its program page, each marked “Paid to the business”: $450K for a 70-person engineering firm’s CAD drawings, calculations, technical reports, revisions and project specifications, and $350K for a 60-person accounting firm’s five years of workpapers, reconciliations, financial models and tax-preparation files. The cards name no companies or dates and are Telegraph’s own statements, not typical amounts. The accounting card includes client tax files, which carry their own rules (accounting firms).

How much is my company’s data worth? sets out the published ranges for whole companies. For a buyer, describe your drives by gigabytes, file counts by type, the years covered, and which folders stay out.