AI data suppliers are increasingly seeking internal emails, chat messages, documents, support tickets and other records that show how real work gets done. The material is valuable because it captures messy professional workflows that are difficult to reproduce with experts performing artificial tasks.
WIRED reports that brokers including Handshake, Mercor and Micro1 are buying or licensing professional data for model training and testing. Offers can cover individual work products or entire corporate archives. A disputed sale of Spirit Airlines’ digital assets included 80,000 email accounts, 500 million internal messages and 20.6 million SharePoint files; the airline’s flight-attendant union objected that employee information was involved.
Brokers say they remove personally identifiable information before data reaches models. Privacy advocates warn that free-form emails and chats can include medical, family or workplace details that are difficult to scrub reliably. Anthropic acknowledged a large market for data but told WIRED it does not buy workflow records from live companies; OpenAI did not respond before publication.
The unresolved issue is ownership: a company may control its systems, but workers may never have agreed to turn their communications into training material.