What changed
AI companies such as Anthropic and OpenAI are paying media companies like Wired between $350,000 and $1.8 million as a one-time fee for access to their historical work records, including documents, emails, Slack messages, and tickets. Some companies are offering up to $2 million for similar data, including call transcripts. This practice is part of a broader industry boom in data acquisition, where companies are buying workflows from sizable organizations to train AI models.
What this means in practice
The companies involved in this data acquisition include startups like Handshake, Micro1, and Mercor, which are snapping up databases as if in a modern gold rush. Handshake, for instance, licenses professional documents across various fields, including journalism, and has been buying graduate students' papers and coursework. Mercor, founded in 2023, claims to have partnerships with major AI labs and is currently the industry leader, with its cofounders becoming billionaires at 22. Micro1 CEO Ali Ansari predicts the data-buying market will grow from $40 to $50 billion annually to over a trillion dollars in the next two to four years.
Limitations and availability
The supplied announcement does not specify the exact number of companies involved in this data acquisition or the exact number of data sets being purchased. Privacy concerns are addressed by companies like Mercor, which claim to strip out personal identifying information (PII) from data before sharing it with AI labs. However, critics argue that anonymizing data may not fully resolve privacy issues, as unstructured data like emails and messages can still contain sensitive information. The sale of Spirit Airlines' digital assets, which included 80,000 email accounts and 500 million internal messages, highlights the contentious nature of this practice, with the flight attendants’ union objecting to the sale of private information.
