AI companies are reportedly offering substantial sums, ranging from hundreds of thousands to millions of dollars, to media companies and other organizations for access to their internal work files, emails, and messages. This data is crucial for training and improving artificial intelligence models.
Several individuals and companies involved in data acquisition have reached out to offer significant fees for access to corporate workflows. One pitch mentioned payments between $350,000 and $1.8 million to media companies for their historical records, including documents, emails, and Slack messages. Another offer hinted at a potential $2 million payout, emphasizing that "public data is exhausted" and AI labs are investing heavily in private data sources.
This burgeoning industry is acquiring vast amounts of professional data, including graduate student papers and coursework. Companies like Handshake and Sunset specialize in brokering these deals, with Handshake confirming they license professional documents to help AI labs test and train models on real-world knowledge work. Sunset focuses on managing the entire process for companies winding down operations, including scrubbing and anonymizing data before selling it to AI labs.
Industry leaders like Mercor, Micro1, and Handshake are aggressively acquiring data. Mercor, founded in 2023, has already seen its young co-founders become billionaires, with its CEO, Brendan Foody, explaining that the demand for high-quality data is immense. Foody asserts that data acquisition startups strip out personally identifiable information (PII) to address privacy concerns, a sentiment echoed by others in the field.
The market for AI training data is projected to explode, with Micro1 CEO Ali Ansari estimating it could surpass a trillion dollars annually within a few years, dwarfing the current $40-$50 billion spent on data acquisition. This quest for data has not been without controversy. A legal battle is underway regarding the sale of Spirit Airlines' digital assets, including millions of emails and messages, after the flight attendants' union raised concerns that the sale involved private employee information.
Critics argue that the anonymization process may not be foolproof, especially with unstructured data like emails and messages, potentially leading to privacy breaches. Furthermore, there's an ethical debate about enlisting workers to contribute data that ultimately fuels AI development aimed at automating the very jobs they perform. While companies like Mercor state their motto is "Organizing human intelligence to power the AI economy," the underlying implication is a potential reduction in work for human employees.
