Email archives for AI training
Exported mailboxes or journaled email with headers, thread structure and attachments metadata, usually from Microsoft 365 or Google Workspace.
- Measured in
- Messages, threads, mailboxes and years covered. Attachments are counted separately.
- Who buys
- Labs and agent builders training models to draft, triage and act on work email; enterprise AI vendors.
- Where it comes from
- Operating companies, closing or bankrupt companies, and estates of acquired businesses.
- Rights to clear
- Employee notices and handbooks, customer and partner confidentiality, attorney-client privilege and any collective bargaining agreement.
- Privacy
- Names, signatures, phone numbers, addresses and anything in the body that identifies a person; health and payroll content excluded outright.
- Pricing signal
- Spirit Airlines: $10 million for about 100 million emails, 500 million Teams messages and other records (August 2026). Most deals are far smaller and priced case by case.
- Business email archives show how real work gets done: requests, negotiations, approvals and follow-ups across long threads.
- In August 2026 Google paid $10 million at a bankruptcy auction for Spirit Airlines data that included about 100 million emails and 500 million Teams messages.
- Privileged legal advice, HR files, health information and third-party confidential material are normally excluded before an archive is offered.
- Email is never licensed as live mailbox access. Buyers receive a prepared, de-identified export.
Why AI teams want email
Email is the largest written record of how organizations coordinate. Models that draft replies, summarize threads or act as agents inside a mailbox learn best from real threads with their full context: who asked, who answered, what happened next. Synthetic email rarely captures that messiness.
What a sale looked like in practice
In August 2026, Alphabet won a bankruptcy auction for Spirit Airlines data with a $10 million bid. The package reportedly included about 100 million emails and about 500 million Microsoft Teams chats, while the customer list and credit card information were excluded (TravelPulse). MIT Technology Review reported the same deal alongside a wider pattern of AI labs buying workspace data from closed startups (MIT Technology Review).
What gets excluded
Lawyers advising on these sales consistently recommend removing high-risk material before any offer (Frankfurt Kurnit Klein & Selz):
- Privileged communications with lawyers, since a sale can waive privilege
- HR, payroll, benefits and health information
- Union communications and anything covered by a collective bargaining agreement
- Third-party confidential material covered by NDAs
- Data from regulated systems with their own rules
Employee communications about pay or working conditions can also be protected activity under the National Labor Relations Act, which is another reason to de-identify thoroughly.
How the archive is prepared
- Inventory mailboxes, date ranges and volumes without exporting anything.
- Agree an exclusion list (people, folders, keywords and domains).
- Run de-identification with an approved provider and keep the protocol on file.
- Produce a small sample for the buyer to evaluate under NDA.
- Deliver the final export by secure transfer once the license is signed.
Pricing reality
The Spirit price works out to a fraction of a cent per message, and that is for a very large, multi-year archive bought in an auction. Smaller companies should expect value to depend on domain, thread quality and how much needs excluding, not on message count alone. Read how to value a dataset.
Questions
Can a company sell its employees' emails?
No federal law specifically bans it, but employee notices, state privacy laws, labor law and confidentiality obligations all apply. Most buyers will only take email that has been de-identified and stripped of HR, health and privileged material.
Does the buyer get access to our email system?
No. A properly structured deal never gives live access. You or your provider prepare an export, and only that export is transferred after the license is signed.
Can a closed company still sell its email archive?
Often yes. The sale is usually run by the people winding the company down, such as a trustee or assignee, and the same exclusions apply. See our guide on selling data from a closed company.
Are attachments included?
Only if agreed. Attachments often hold the most sensitive material, such as contracts and spreadsheets, so they are usually reviewed and licensed separately.
Related guides
Related data types
Sources
Need email archives, or hold it?
Tell us what you need or what you hold. We check rights and privacy first, then introduce vetted counterparties. You license directly with each other.
- We reply to every enquiry within two working days.
- No upfront fees for sellers
- We never hold your raw data