AI firms aggressively acquire internal corporate data for training

Major artificial intelligence companies and data brokers are increasingly targeting internal corporate records, including emails, Slack messages, and proprietary code, to refine their AI models. The industry shift toward acquiring private operational data has intensified throughout 2026, as developers seek to understand how work is performed in real-world environments.

Financial incentives for companies to share their internal files are significant. Businesses are receiving offers ranging from $100,000 to $2 million for their data. According to Gartner estimates, annual spending on data for AI training is projected to grow from less than $1 billion in 2025 to nearly $6.5 billion by 2027.

The acquisition process is often characterized by high-pressure tactics. Ryan Lock, cofounder of the brokerage firm Polyshares, stated that some companies are offered double the market rate with a requirement to accept the deal within 48 hours. Lock compared the current market environment to the early stages of the Texas oil boom, while Mode.Inc CEO Dan Novaes described the demand for this data as astronomical.

Industry leaders argue that this information is essential for technological advancement. Ali Ansari, CEO of Micro1, noted that noise in real-world environments is required to accurately test models. Major entities identified as seeking this data include OpenAI, Anthropic, Meta, Google DeepMind, and xAI, often working through intermediaries like Micro1, Polyshares, and Troveo.

The practice has raised concerns regarding employee privacy, as workers are often unaware that their internal communications are being sold. Some organizations have faced internal pushback, including instances at Meta where employees resisted plans to harvest their data. As AI labs shift their focus from public internet data to long-horizon records that track the progression of work, the market for private corporate files is expected to continue its expansion.