Licensing a Company Email Archive for AI Training (Not Selling Email Lists)

A company email archive can be licensed to AI developers as training data under a contract, which is a different trade from selling lists of email addresses for marketing. The only court-filed count is in the Spirit Airlines bankruptcy: 100,000,000 emails across 80,000 email accounts, inside a bid for the whole schedule that the court had not approved as of November 9, 2026.

An email archive is one record type in AI training data licensing, and not licensing it stays a valid result.

Program terms were checked on November 9, 2026 and the court record and mailbox rules on November 10, 2026 by DataPacts Editorial Team: how we source, date and check every figure.

What an Email Archive License Covers, and Why Email Lists Are Not Part of It

An email archive license grants an AI developer the right to train on a defined set of mailboxes, threads and attachments for a defined use, while the company keeps the original mail. micro1's Data License Master Terms (updated November 8, 2026) describe a worldwide, royalty-free, nonexclusive, irrevocable, perpetual license. SuperAnnotate states that the seller is "licensing rights for a set period, not selling the data" (checked November 9, 2026).

Address lists sold for marketing are outside this reference.

Which AI Data Programs Ask for Email, Calendar and Mailbox Records

Twelve AI data programs name email, Gmail, Outlook or calendars among the records they license, each on its own page checked November 9, 2026. Most of them run company data partnership programs that connect to the mail system or export an agreed copy.

The table compares what each program says it acquires, its email-relevant terms and its published figures, as of 2026-11-09, in alphabetical order.

Three programs contradict themselves on their own pages, and both claims stay visible. Handshake AI shows "Earn $100K to $4M+" next to an estimator returning "up to $100,000", and Polyshares shows "Defined material for a defined term" beside a "Perpetual exclusive license" (both checked November 9, 2026). Troveo, a marketplace licensing email and calendar records on the seller's behalf, states "$20M+ Paid to Licensors" and "$50M+ paid to creators globally" on two pages (read November 10, 2026).

Brokers such as Avelence, Data Licensing Advisors, OpenTrain AI, Pewter and SimpleClosure AssetHub also list email (checked November 9, 2026). None acquires data; they place it, as set out under AI data licensing brokers and sell-side advisers.

What Gives an Email Archive Value to a Buyer

Email is valued for the threads that tie a request to its outcome across years, and programs price it as part of a company's whole record set rather than per message. Mercor states that "Connecting more tools generally results in a higher payout", and AfterQuery names years of history and connected systems as price factors (both checked November 9, 2026).

No public figure exists for a price per email or per mailbox. The court-filed count of 100,000,000 emails sits inside one bid for the whole Spirit Airlines schedule, so no part of that bid can be assigned to email.

Programs quote ranges for a whole record set, each its own claim: Scale AI "$10K–$1M+ illustrative", Pangea.ai "around $100K to $2M" and Mode "$100K–$5M" (checked November 9, 2026). The methods for estimating what company data is worth to AI companies cover such ranges.

Whose Words Sit in a Mailbox: Staff, Former Staff and Outside Correspondents

A mailbox holds mail written by current staff, departed staff and outside correspondents, and the company's rights differ for each group.

In the United States, 17 U.S.C. 201(b) makes the employer the author of a work made for hire, defined in 17 U.S.C. 101 as a work prepared by an employee within the scope of employment. In the United Kingdom, section 11(2) of the Copyright, Designs and Patents Act 1988 makes the employer first owner of a literary work made by an employee in the course of employment. Both reach only works made by employees, so neither makes the company author of incoming mail.

For departed staff, section 26(8) BDSG counts persons whose employment has ended as employees (section 26(1) is disapplied by BAG 8 AZR 209/21 of May 8, 2025). The steps for employee notice and consent when licensing internal communications sit on a separate page.

Mailbox Rules That Go Beyond General Employee Data Law

Four rules attach to mailboxes as such, on top of general data protection and employee-data law, and each limits what a company may hand to a third party. The table lists them as of 2026-11-10.

Correspondence Secrecy Where Staff Could Use Work Mail Privately (Germany)

Under section 3 TDDDG (in force December 1, 2021, renamed May 14, 2024), a telecommunications provider may pass message content to others only where a statute expressly allows it, and older German practice treated employers who allowed private email as such providers. The Federal Network Agency (July 2025) and the North Rhine-Westphalia commissioner (29th report, 2024) say employers are generally not bound; GDPR still applies.

Accounts of Departed Staff Kept Active (Italy)

Italy's data protection authority fined a company in liquidation EUR 5,000 on April 17, 2026 for keeping a former employee's two mailboxes active for six months. The order (registro n. 278) records that the company kept them beyond its own three-month period without a balancing test.

Monitoring Rights That Authorize No Disclosure (Poland and Italy)

Polish and Italian labor law let an employer monitor or use work mail for stated employment purposes, and neither provision names disclosure to an outside party. Article 22³ §1-3 of the Polish Labor Code (consolidated text of September 24, 2026) allows monitoring business email where necessary. Article 4(3) of Italy's Law 300/1970 makes monitoring information usable for employment purposes after adequate notice.

Provider Disclosure Limits Under 18 U.S.C. 2702 (United States)

18 U.S.C. 2702 bars providers of communication services to the public from divulging stored content without lawful consent, and a company's internal mail system is generally not a service to the public. Section 2702(b)(3) excepts lawful consent of the originator, an addressee or an intended recipient.

Getting Mail Out: Admin Exports, Connector Apps and the API Rule

Mail leaves a hosted system either through the tenant's own admin export or through a connector app, and the route decides which vendor rule applies. Google's Workspace API user data policy prohibits using API user data to train or improve an AI model, beyond a user's personalized model. Google's admin Data Export has its own clocks: a super admin account at least 30 days old and availability from 48 hours to 14 days. A Microsoft Purview eDiscovery export needs Enterprise E3 or E5 and expires after 14 days (all checked November 9, 2026).

Mercor and Centific describe connector routes. Steps for each mail system sit under export rights and mechanics by source system, including mail.

What Comes Out of an Email Archive Before Delivery

Before delivery, the seller chooses which mailboxes, teams and periods go in, and buyers' own pages name categories that never go in. Centific enforces exclusions in its extraction code, and AfterQuery never uses payroll, privileged legal material or credentials (both checked November 9, 2026).

Ooak Data annexes exclusions to the contract and deletes the raw archive within 30 days of delivery. Scale AI anonymizes inside the company's environment before extraction, and Mercor states that "No AI lab buyer or third-party vendor ever sees your raw data" (checked November 9, 2026).

Headers, signatures, quoted replies and attachments carry names and addresses that a body-text filter misses. Methods for PII removal and de-identification before licensing handle those fields.

An Email Archive in a Court-Supervised Sale

The Spirit Airlines bankruptcy is the one court record that counts an email archive offered to AI buyers: 100,000,000 emails across 80,000 email accounts, listed as Included. The schedule (Dkt. 1463, Exh. A-2, August 14, 2026) also lists 500,000,000 Teams items as Included.

The debtors named Google LLC successful bidder at USD 10,000,000 and Mercor.io Corporation alternate bidder at USD 7,500,000 for the whole schedule (Dkt. 1463, August 14, 2026). The court had not approved the sale as of November 9, 2026, and the hearing is adjourned to November 17, 2026 (Dkt. 1779).

An independent Deidentification Agent, Tonic.ai, handles unstructured data such as corporate emails under the Cal. Civ. Code 1798.140(m) standard that the debtors' letter of August 5, 2026 adopted. The ombudsman's supplemental report (Dkt. 1684, October 5, 2026) says approval of the de-identified data sale would be appropriate from a privacy perspective, and footnote 3 states that risks to employees and third parties were not investigated.

Licensing the Archive, a Subset or Nothing: Benefits, Risks, Downsides and Alternatives

Licensing an email archive brings income from a record the company already holds and exposes it to the people who wrote the mail, so not licensing stays a valid result.

Benefits

• Mercor states that the wire typically follows 2 to 4 weeks after the signed agreement (checked November 9, 2026).
• Centific lets the owner choose applications, teams and periods.
• SuperAnnotate lets the seller approve each buyer.

Risks

• In the Spirit record, risks to employees and third parties were not investigated.
• Italy's regulator fined a company for a departed employee's mailboxes kept active (EUR 5,000, April 17, 2026).
• Google's API rule bars AI training with API user data, so the route can end a plan.

Downsides

• micro1's license is irrevocable and perpetual once it attaches, so a grant cannot be recalled.
• Deletion terms differ: Ooak Data within 30 days of delivery, micro1 at the earlier of 120 days after delivery or termination.
• No public price per message exists to benchmark an offer.

Alternatives

• License a subset: chosen teams or periods, as Centific and AfterQuery allow.
• License other record types first and leave mail out.
• Do not license and keep the archive under the company's retention rules.

Each path is weighed in the wider question of should you sell or license your data.

Where an Email Archive Sits Among Workplace Data Types and the Seller Process

Email is one of the workplace record types AI programs license, next to chat, documents, tickets and code, and it passes through the same seller steps. The parent page covers every data type AI companies buy, and chat sits next to mail in selling chat messages (Slack, Teams) for AI training.

Sibling data types:

• Source code and codebases
• Documents, wikis and drive files
• Issue trackers and service-desk tickets
• Call recordings and sales calls

Seller steps that apply to an email archive:

1. Data sale readiness
2. Rights audit
3. Customer consent and confidentiality clauses
4. The data manifest
5. Choosing a buyer for your data
6. AI data licensing agreements

Email Archives From a Closed or Insolvent Company

Mail from a closed company has a deletion clock: Microsoft 365 for business deletes data no later than 180 days after cancellation (page of June 8, 2026), and a Google Workspace export excludes core service data once the subscription is cancelled (checked November 9, 2026). See the steps for selling data from a shut-down or defunct company. Another explains how trustees sell data assets in insolvency; the Spirit record is under Spirit Airlines' data archive sale.

Mailbox Rules by Country and Export Guides by Mail System

This page is information, not legal advice.

• Employee data and AI training rules by jurisdiction
• GDPR and AI training data
• Germany: telecom secrecy and private use of work email
• Garante privacy decisions and employee mailboxes
• Licensing employee email and chat data in Italy
• Poland: rules for licensing data
• ECPA, the Wiretap Act and stored communications
• Google Workspace (Gmail, Drive) exports
• Microsoft 365 (Outlook, SharePoint, OneDrive) exports

Buyer Profiles Cited on This Page

The program terms above come from each company's own pages, checked November 9, 2026.

• AfterQuery
• Centific
• Collinear AI
• Handshake AI
• Mercor
• micro1
• Mode
• Nyne
• Ooak Data
• Pangea.ai
• Polyshares
• Scale AI
• SuperAnnotate
• Troveo

Compare who buys AI training data and AI data partnership payouts and program terms compared.

Frequently Asked Questions

Can a company sell its old emails to AI companies?

A company usually licenses its email archive rather than selling it, granting a right to train while keeping the mail. micro1's terms describe a nonexclusive license.

How much are a company's emails worth to AI buyers?

No public price per email or per mailbox exists; programs quote ranges for a company's whole record set, each as their own claim. Scale AI states "$10K–$1M+ illustrative" (checked November 9, 2026).

Does Google train its AI models on a company's Gmail?

Google states that Workspace does not use customer data to train models without the customer's prior permission or instruction (privacy hub, last updated October 26, 2026).

Consumer settings are covered under opting out of AI training.