A growing number of failed startups are finding unexpected value in their shutdown process: selling the digital trail of their day-to-day operations to artificial intelligence companies hungry for new training data.
According to a report from Forbes, this emerging practice is turning what was once considered useless corporate residue into a potentially lucrative asset. When Shanna Johnson wound down the transcription and captioning firm cielo24, she discovered that years of internal communications—ranging from Slack messages and emails to IT tickets and shared documents—could be packaged and sold. The company ultimately generated hundreds of thousands of dollars by allowing its internal data to be used for AI training.
A New Kind of Asset
The process was facilitated by SimpleClosure, a firm that typically handles administrative tasks like payroll termination and regulatory filings. Increasingly, however, it is also helping companies monetize their “operational exhaust”—a term used to describe the vast amount of data generated through routine business activity.
This shift reflects a broader trend in the AI industry. Developers are seeking more realistic and detailed datasets as traditional sources—such as publicly available web content—become less useful for advancing next-generation systems.
Ilya Sutskever has suggested that much of the easily accessible internet data has already been consumed for training purposes. More importantly, such data often lacks the complexity needed to train AI systems designed to perform real-world tasks, such as managing workflows or making business decisions.
Demand for “Real-World” Data
That gap has fueled demand for internal company records, which capture how organizations actually function. These datasets include messy, unstructured, and highly contextual information—precisely the kind of material AI developers now consider valuable.
Ali Ansari, whose company micro1 develops training environments for AI agents, has pointed to the importance of this realism. His firm’s “Roots” platform simulates business operations, allowing AI systems to practice tasks like financial management and scheduling in environments that resemble real workplaces.
SimpleClosure CEO Dori Yona described demand from AI companies as intense, likening it to a “gold rush” for high-quality, real-world data. The company is now building a marketplace called Asset Hub, where firms in the process of shutting down can sell codebases, communication logs, and other digital assets. Over the past year, SimpleClosure has reportedly completed nearly 100 such transactions, generating over $1 million in returns for founders.
Privacy and Ethical Concerns
The trend, however, raises complex questions about privacy and consent. Internal communications often include identifiable information about employees, even if companies attempt to remove sensitive data before selling it.
Marc Rotenberg, founder of the Center for AI and Digital Policy, has warned that employees may not have anticipated their workplace messages being repurposed for AI training. While many workers sign agreements granting employers rights over work-related materials, the downstream use of that data in AI systems introduces new ethical considerations.
Rotenberg’s organization has called for greater oversight, including potential scrutiny from regulators, arguing that existing safeguards may not adequately address the risks associated with repurposing internal communications.
A Shifting AI Landscape
The monetization of “operational exhaust” highlights a significant shift in how data is valued in the AI economy. As developers push toward more advanced, task-oriented systems, the need for real-world training material is reshaping what counts as a valuable asset—even in failure.
For companies on the brink of closure, this development offers a rare upside. But for employees, regulators, and the broader public, it opens a new front in the ongoing debate over data ownership, privacy, and the future of artificial intelligence.
READ NEXT: Anthropic’s AI Blackmail Hoax Is A Play For Regulatory Capture
