Tech companies are buying internal corporate files and emails to train artificial intelligence models. Data brokers now act as middlemen in these deals, sourcing material from businesses of all sizes.
The demand for real-world work documents has grown sharply as AI firms seek better training data. Generic web content no longer offers enough depth for advanced models. Internal files provide the context and structure that public sources lack.
Emails, spreadsheets, project plans, and meeting notes are among the materials in demand. Data brokers collect these files from companies willing to sell. Buyers then use the content to improve how AI systems understand workplace tasks.
Not everyone involved is comfortable with the practice. Some employees worry about private conversations and sensitive information changing hands. Others question whether they consented to such use of their work.
Companies selling the data often see it as a way to monetize unused archives. Legal teams, however, must weigh privacy rules and contractual obligations. The risk of exposing trade secrets or personal details remains a central concern.
The trend reflects a broader scramble for high-quality training data. AI developers have already scoured public websites, books, and forums. Work files represent one of the few remaining large sources of authentic human writing.
Regulators have begun to take notice of the emerging market. Questions about consent, ownership, and data protection are likely to shape how these deals proceed. For now, the buying continues, with little public visibility into what changes hands.
