A growing number of failed startups are turning their internal workplace communications into a lucrative final asset, quietly selling years of Slack messages, emails, and project records to artificial intelligence firms for as much as $100,000.
The emerging trend underscores the intensifying global demand for real-world data to train advanced AI systems, with transactions being facilitated by SimpleClosure, a company that helps startups shut down operations and recover value from remaining assets.
In the past year alone, SimpleClosure has reportedly handled nearly 100 such deals, connecting defunct companies with AI labs seeking authentic workplace data. These archives often include unfiltered conversations, internal debates, brainstorming sessions, and operational workflows — material now considered highly valuable in building the next generation of intelligent digital assistants.
Speaking to Forbes, the company’s CEO, Dori Yona, described demand from AI buyers as “insane,” noting that the industry is experiencing a “gold rush” for high-quality, human-generated data.
Among the most notable cases is Shanna Johnson, former CEO of cielo24, who sold her company’s entire 13-year digital archive after it shut down. The dataset included everything from Slack conversations and emails to project management logs. According to Johnson, the proceeds helped the company conclude its operations cleanly amid financial challenges.
AI developers are using such datasets to create simulated work environments, often referred to as “reinforcement learning gyms,” where AI agents are trained to perform real-world office tasks such as writing reports, scheduling meetings, and managing team collaboration. Unlike synthetic datasets, these archives capture the messy, nuanced realities of human communication, making them especially valuable for training systems designed to function as digital coworkers.
However, the practice is raising serious privacy concerns. Critics argue that employees were never informed that their internal communications could eventually be sold for AI training purposes. Even when companies claim to remove personally identifiable information, experts warn that contextual clues — such as project details, writing styles, and internal references — may still make individuals identifiable.
Privacy advocate Marc Rotenberg of the Center for AI and Digital Policy has previously warned that workplace communication tools blur the boundary between private expression and corporate data ownership, leaving employees exposed when companies dissolve.
Under current legal frameworks, ownership of internal company data typically transfers to the company or its liquidators after shutdown, giving former employees little or no control over how their communications are used or sold.
While SimpleClosure says it is refining processes to remove sensitive information before any sale, the development has intensified calls for tighter regulation and clearer consent policies around data usage.
For struggling founders, the rise of this secondary market offers a rare opportunity to recover value from failed ventures. But for millions of workers whose casual office conversations may now be repurposed to train AI systems, it raises an unsettling question about digital ownership in the modern workplace.
As the AI industry’s appetite for authentic human data continues to surge, what was once dismissed as routine office chatter may now carry unexpected value — and consequences — long after a company shuts its doors.
