Tech Startups Profit by Selling Employee Slack and Email Data to Train AI Models

There’s a new link in the food chain of tech startups: founders are monetizing the digital remains of their defunct companies by selling employees’ Slack messages and emails as training data for AI models. Forbes reports that a burgeoning ecosystem of intermediaries has emerged to facilitate these transactions.

Quality training data is critical for building AI models, yet sources are increasingly scarce in a post-AI-ransacked internet. Data from digital workplaces like Slack is particularly valuable because it enables AI researchers to create realistic sandboxes for training agents that perform work tasks. “Model companies are realizing the noise in the real-world environments is required to accurately test models,” said Ali Ansari, whose company micro1 sells a mock holding company for AI agents to learn in.

Reinforcement Learning Gyms: A Nascent Industry

These simulated workplace environments are known as reinforcement learning gyms, or “RL gyms,” and their creation has become its own industry. Demand is surging: Anthropic is reportedly considering spending $1 billion on RL gyms this year, according to The Information. Multiple RL gym startups, including Prime Intellect and Fleet, are now valued at similar figures.

Middlemen Facilitate the Sale of Digital Assets

Founders of defunct startups are eager to sell their digital assets, and middlemen have stepped in to streamline the process. SimpleClosure, which brands itself as the “TurboTax of shutting down,” recently launched Asset Hub, a tool that allows moribund companies to sell Slack archives, emails, and code libraries. The company claims to anonymize the data before it reaches buyers.

SimpleClosure’s CEO, Dori Yona, told Forbes that the company has processed nearly 100 deals for defunct companies in the past year, recovering over $1 million for their founders.

Ethical and Privacy Concerns Linger

Despite the financial incentives, the practice raises significant ethical and privacy questions.

“I think the privacy issues here are quite substantial. Employee privacy remains a key concern, particularly because people have become so dependent on these new internal messaging tools like Slack.”

“It’s not generic data. It’s identifiable people.”

Experts also question the effectiveness of anonymization in the data being sold. “If anonymization’s not done correctly, there are risks that companies who have access to the data would be able to,” Roteberg added, though his statement was cut off in the original report.