AI firms aggressively pursuing corporate and media data for model training

Major artificial intelligence laboratories, including Anthropic and OpenAI, are reportedly paying media organizations substantial sums, ranging from $350K to $1.8M, for access to historical proprietary data such as emails and documents. Some pitches have escalated to potential $2 million offers, driven by the industry's reported spending of over $45B on private data to advance model capabilities. This data acquisition boom involves specialized startups like Handshake, which offered a freelance writer $1,000 for her work, and Sunset, which assists companies winding down by anonymizing their data for AI labs. Industry leaders like Mercor, founded in 2023, are rapidly acquiring databases, with its cofounders becoming billionaires last year. Mercor CEO Brendan Foody stated the market is set for explosive growth, emphasizing that data gathering is as crucial as compute power. Micro1 CEO Ali Ansari projects the annual data market could surpass a trillion dollars within two to four years. While Mercor claims data is de-identified to protect privacy, critics like Mayu Tobin-Miyaji of the Electronic Privacy Information Center caution that unstructured data in communications may still pose privacy risks. The practice has drawn scrutiny, notably during a Spirit Airlines asset sale, where a flight attendants’ union objected over the inclusion of private information.