AI labs are offering companies between $350,000 and $2 million for one-time access to their internal work records—including documents, emails, Slack messages, support tickets, and call transcripts—to train better AI models, according to a report published by WIRED in its Backchannel newsletter. The investigation uncovered a rapidly expanding data acquisition industry that's quietly purchasing workplace workflows from businesses of all sizes, with startups flooding inboxes to broker deals between struggling companies and major AI labs. The practice has become so widespread that it's begun to raise serious questions about worker privacy and the ethics of using employee communications to build automation technology designed to replace those same workers.
The data-buying market spans multiple industries and worker categories, the report finds. Handshake, one of the leading companies in this space, confirmed it offered a freelance writer $1,000 for her work files on a single story, and told WIRED that journalism is among many professional fields it targets. Graduate students' papers and coursework represent a significant portion of Handshake's business, according to the company. Meanwhile, startups are receiving cold emails from firms like Sunset that promise to "handle the full process" of shutting down companies while "scrubbing and anonymizing the data" before selling it to AI labs. The most dramatic example came when bankrupt Spirit Airlines put its digital assets up for auction—80,000 email accounts, 500 million internal messages, 20.6 million Microsoft SharePoint files, and a million time-card records—with Google bidding $10 million before Micro1 topped the offer at $12.5 million, though the sale is now on hold following privacy objections from the flight attendants' union.
Industry leaders in this emerging sector have achieved explosive growth in just a few years. Mercor, founded in 2023, saw its three cofounders become billionaires at age 22 last year, according to the report. Mercor CEO Brendan Foody explained to WIRED that AI labs were frustrated by data quality from hired experts simulating workflow, which didn't capture "the messiness of an actual operating business." All the data acquisition firms claim to strip out personally identifying information before selling to labs, with executives insisting technology has solved privacy concerns. However, Mayu Tobin-Miyaji, general counsel of the Electronic Privacy Information Center, which filed an amicus brief for the Spirit Airlines flight attendants, countered that "emails and Teams messages contain unstructured data, which is harder to root out," noting that employees discuss doctor's appointments and family matters in workplace communications.
The market for this data is poised to explode dramatically, industry executives say. Micro1 CEO Ali Ansari told WIRED that labs, enterprises, and robotics companies currently spend about $40 to $50 billion annually acquiring data, and predicted "in the next two to four years, this is going to be over a trillion dollars a year." Mercor's Foody urged the publication to consider that Big AI has two bottomless needs—compute and data—and while the industry publicly invests heavily in the former, data gathering is equally critical. When asked whether corporations will eventually sell their workflow data to AI companies on an ongoing basis, Ansari answered yes, predicting "it's going to become more common than ads as a revenue line item for companies." Venture capitalist Gordon Ritter, who has taken on the issue as a cause, warned that "the AI industry is moving far faster than our ability to actually set the standards, and it's partially intentional." The most troubling dimension, the report notes, is that workers are essentially being enlisted—often without explicit consent—to fuel an effort whose ultimate goal is automating or replacing the human labor that created those documents, messages, and emails in the first place. The workforce commodification raises stakes far beyond individual privacy breaches, suggesting a structural realignment in how labor itself becomes raw material for its own obsolescence.

