Technology news and discussion on TalkFuse.
AI Companies Really, Really Want to Buy Your Work Files and Emails
AI companies are increasingly paying for private work files, emails and other business records to train and evaluate models, creating a fast-growing market that has received less attention than spending on data centres. The trade matters because real workplace material can capture the complexity of day-to-day tasks, while raising questions about consent, privacy and how companies’ information is used.
Data brokers including Handshake, Micro1 and Mercor say they buy or license material such as professional documents, student coursework and company workflows. One writer was offered $1,000 for a story, while a separate pitch suggested a deal worth up to $2 million; Micro1’s CEO estimated current data spending at $40 billion to $50 billion annually, a figure the article says it could not verify. Brokers say they remove identifying details, but AI firms gave limited confirmation of their purchasing practices, and forecasts of spending exceeding $1 trillion a year within two to four years remain industry claims.
In short
- AI firms are buying private work data to improve models.
- Brokers say the market may grow rapidly.
- Privacy and consent remain key concerns.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Paying for work files and emails could give AI developers realistic examples of how people handle complex tasks, making models more useful and reliable than systems trained on less representative material. Licensing this data can also compensate contributors and businesses for valuable work, while anonymisation and clear agreements may help manage privacy risks.
The case against
Work files and emails often contain sensitive information about employees, clients and organisations, and removing names may not prevent people or details from being identified. People whose information appears in the material may not have agreed to its use, while limited transparency about buyers and handling practices makes it difficult to assess whether safeguards and compensation are adequate.