A burgeoning industry is emerging where artificial intelligence labs are acquiring data from defunct startups to enhance their AI models. This trend sees the digital archives of failed companies, including codebases, emails, chats, and documents, being sold to AI developers seeking valuable training material.
A New Market for Startup Data
Companies like SimpleClosure and Sunset are facilitating these sales, acting as intermediaries for startups winding down operations. SimpleClosure's Asset Hub platform has already brokered over 150 deals, with projections to return over $10 million to startups by the end of 2026. CEO Dori Yona notes that these transactions, once considered taboo, are becoming routine, with 47% of SimpleClosure deals now involving code or data sales. Sunset, another player in this space, values client assets based on factors like headcount, data volume, and team tool usage, likening the operational data to valuable 'wood shavings'.
Driving Demand for Real-World Data
The demand for this data surged last year as AI training labs encountered limitations with public and synthetic datasets. Surge AI, for instance, uses acquired data to build 'reinforcement-learning gyms' for AI agents. Nick Heiner of Surge AI emphasizes the value of data archives that capture a wide range of situations, noting that a company's Slack history, if rich in diverse conversations, can be more valuable than combined data from multiple sources like Slack, email, and calendars.
Legal and Ethical Considerations
While the market is expanding, legal and ethical questions persist. The sale of data, such as Spirit Airlines' extensive archives including emails and employee records for $10 million, has faced objections on privacy grounds. Attorneys have described such data as 'legally radioactive,' despite efforts by suppliers and buyers to scrub personally identifiable information. Concerns remain about private details lurking in communications, highlighting a 'Wild West' phase that may evolve towards more formal licensing agreements.
💬 Comments (0)