Table of Contents
Artificial intelligence companies are aggressively seeking to purchase private work files, emails, and call transcripts to train their models. As public data sources become exhausted, these firms are spending billions of dollars to acquire proprietary workflows, often reaching out directly to individuals and companies to secure access to internal information.
The Booming Market for Private Data
Reporting for WIRED, Steven Levy notes that the industry for data acquisition has grown significantly, with companies like Handshake, Micro1, and Mercor leading the charge. These firms act as brokers, purchasing professional documents, graduate coursework, and even the digital assets of bankrupt companies. Handshake, for instance, licenses professional documents across various fields, including journalism, to help labs test how models handle real-world knowledge work.
Other startups, such as one named Sunset, specialize in helping failing companies wind down by scrubbing and anonymizing their internal data before selling it to AI labs. Mercor, founded in 2023, has emerged as a major player in this space. Its CEO, Brendan Foody, explains that the industry is driven by a need for high-quality, “messy” data that reflects actual business operations, which is superior to simulated workflows.
Privacy Concerns and Legal Challenges
While data brokers claim that they strip out personal identifying information (PII) before handing data over to AI labs, critics remain skeptical. The Electronic Privacy Information Center (EPIC) has raised alarms about the difficulty of fully anonymizing unstructured data, such as emails and internal messaging platforms, which may contain sensitive personal details about employees.
The risks of this practice were highlighted during the bankruptcy proceedings of Spirit Airlines. A bidding war erupted between Google and Micro1 over the airline’s digital assets, which included millions of internal messages and SharePoint files. The sale was halted after the flight attendants’ union objected, arguing that the transaction involved the unauthorized sale of their private information. The union president noted that this data is being commodified specifically to train models that could eventually replace the workers themselves.
The Future of Data Acquisition
Industry leaders believe the market for data will continue to expand rapidly. Micro1 CEO Ali Ansari estimates that while companies currently spend $40 to $50 billion annually on data, this figure could reach $1 trillion in the next few years. Ansari suggests that selling workflow data will eventually become a standard revenue stream for corporations, similar to advertising.
Despite the pushback, proponents of the industry argue that the trade-off is justified by the potential for AI to create greater abundance and efficiency. Venture capitalist Gordon Ritter, however, warns that the industry is moving faster than society can establish necessary standards. He argues that the current lack of regulation is partially intentional, allowing AI firms to operate in a “Wild West” environment.
The aggressive nature of these data-gathering efforts was further illustrated when a representative from a startup called Tenkara attempted to solicit editorial archives from WIRED. The company, which aims to automate office operations for manufacturing firms, confirmed that cold-calling to acquire data is a core part of its business strategy.
