AI Labs Are Quietly Acquiring Corporate Data at Unprecedented Scale, From Emails to Slack Messages
A booming data-brokerage industry is selling companies' internal workflows to AI developers for millions, raising questions about worker consent and competitive advantage.
Last verified:
A thriving ecosystem of data brokers is systematically acquiring the internal communications and workflows of companies worldwide and reselling them to artificial intelligence laboratories, according to reporting by Wired AI. The practice, which has accelerated as publicly available training data becomes scarce, involves per-transaction valuations ranging from $350,000 to $2 million and represents a market reportedly worth over $45 billion annually in private-data procurement alone.
The Data Brokerage Boom
According to Wired AI, three companies—Mercor, Handshake, and Micro1—have emerged as industry leaders in acquiring and licensing corporate datasets. Mercor, founded in 2023, began as an AI recruiting platform before pivoting to data acquisition; its three cofounders reportedly became billionaires by age 22. Handshake, which brokers professional documents across multiple sectors including journalism and academia, has offered individual freelancers up to $1,000 per article in exchange for usage rights. A Handshake representative told Wired AI that “We license professional documents across a lot of fields so the labs can test and train how models handle real knowledge work.”
The scope of acquisitions extends beyond individual contributors. According to Wired AI, newer startups like Sunset have begun offering services to companies seeking orderly shutdown procedures, explicitly framing data monetization as part of the wind-down process. The company “scrubs and anonymizes the data so the company people and any identifying details are stripped out, then brokers the deal with the labs,” a Sunset representative explained to the publication.
Scale and Market Drivers
Wired AI reports that AI labs are spending over $45 billion annually on private data acquisition, signaling a strategic pivot away from publicly sourced training material. This shift reflects the sector’s acknowledgment that internet-scale data—long the foundation of large language model training—has been largely exhausted. Mercor CEO Brendan Foody told Wired AI that the market is positioned for explosive growth as AI developers demand greater volumes of specialized datasets.
When contacted for comment, Anthropic acknowledged the private-data market but stated that it does not purchase workflow data from active companies like Wired. OpenAI did not provide a response to Wired AI before publication.
Why This Matters
The normalization of corporate-data acquisition raises unresolved questions about worker consent, data provenance, and competitive risk. Most individual employees have no visibility into—or authority over—whether their employers have monetized their communications or work product. Organizations operating in regulated industries (financial services, healthcare, legal) face unclear compliance exposure if their internal data becomes part of third-party model training without explicit contractual disclosure to clients or regulators. Additionally, the commodification of domain-specific workflows could entrench competitive advantages for AI labs that afford the highest prices, while creating perverse incentives for financially distressed companies to sell irreplaceable institutional knowledge.
Frequently Asked Questions
What types of corporate data are being purchased?
Email archives, internal documents, Slack messages, call transcripts, support tickets, and coursework. Data brokers typically anonymize and scrub identifying information before selling to AI labs.
How much are companies paying for this data?
According to Wired AI, individual deals range from $350K to $2M as one-time fees. Freelancers have been offered $1,000 per story or article.
Which companies are buying this data?
According to the article, data brokers claim partnerships with major AI labs including Anthropic and OpenAI, though both companies were circumspect when contacted for comment.
What happened to Mercor's founders?
The three cofounders of Mercor, founded in 2023, became billionaires by age 22, according to Wired AI.