The $45 Billion Data Gold Rush: How AI Labs Are Buying Your Corporate Emails and Work Files
AI labs are paying millions for corporate emails and work files to train better models, sparking a data gold rush and privacy concerns.
S
··2 min readSciTech Editorial Team
Newsroom

A quiet but massive financial shift is reshaping the artificial intelligence landscape, as major AI labs like Anthropic and OpenAI reportedly pay media companies and enterprises between $350,000 and $2 million for copies of their historical work records. These records include documents, emails, Slack messages, and support tickets, which are used to train more robust and context-aware models. This surge in demand has led to a new industry of data brokers, with some pitches suggesting that public data is exhausted and that labs are spending over $45 billion annually on private data to maintain their competitive edge.
The market for professional workflow data has become a modern gold rush, dominated by startups like Handshake, Micro1, and Mercor. Mercor, founded in 2023, has seen its three cofounders become billionaires at the age of 22, citing partnerships with nearly all major AI labs. Brendan Foody, CEO of Mercor, explains that the industry previously relied on simulated workflows, which failed to capture the inherent messiness of real-world business operations. Consequently, the focus has shifted to acquiring authentic, unstructured data from live companies to bridge the gap between theoretical training and practical application.
Micro1 CEO Ali Ansari predicts that the data acquisition market will soon rival the scale of AI compute infrastructure, potentially exceeding one trillion dollars annually within the next two to four years. He notes that a significant portion of recent Y Combinator startups are now focused on data services, reflecting the intense demand from labs, enterprises, and robotics companies. This trend highlights a dual bottomless need in the AI sector: while compute power receives public attention, data gathering is equally critical for model performance.
However, this data boom has sparked significant privacy controversies, most notably involving the sale of Spirit Airlines’ digital assets. The corpus included 80,000 email accounts and 500 million internal messages, leading to a bidding war between Google and Micro1. The sale is currently on hold after the flight attendants’ union argued that the data contained private information. Mayu Tobin-Miyaji, general counsel of the Electronic Privacy Information Center, warns that anonymization may not fully eliminate privacy risks, as unstructured data like emails often contains sensitive personal details that are difficult to completely scrub.
Despite these concerns, data acquisition firms maintain that their de-identification processes are robust and verifiable. They argue that technology has effectively solved the privacy issues associated with purchasing such vast amounts of information. Nevertheless, the incident involving Spirit Airlines underscores the potential ethical and legal challenges that may arise as the industry continues to expand its reach into the private workflows of millions of workers.



