Shuttered Startups Are Selling Old Slack Chats and Emails to AI Companies

Why It Matters

The practice creates a new revenue stream for failed ventures while feeding AI labs with high‑quality, context‑rich data, accelerating the development of autonomous agents. At the same time, it spotlights significant employee‑privacy risks that could trigger regulatory scrutiny.

Key Takeaways

• Defunct startups sell Slack, email, Jira data for AI training
• Deals average $10k‑$100k; 100 transactions processed last year
• Privacy experts warn anonymized data may still expose personal info
• AI labs use workplace data to build autonomous agent models
• New model creates revenue stream for shuttered companies

Pulse Analysis

The emergence of a data‑sale marketplace for shuttered startups reflects a pragmatic response to the costly reality of corporate failure. By bundling years of internal communications—Slack threads, email chains, and Jira tickets—these firms offer AI researchers a trove of real‑world context that public web data cannot provide. SimpleClosure, a specialist in wind‑down services, has already brokered roughly a hundred deals, each netting between $10,000 and $100,000, and some CEOs claim earnings in the high‑hundreds of thousands. This model not only recoups capital for founders but also fuels the rapid iteration of large‑language and agentic models that require nuanced, domain‑specific inputs.

However, the privacy implications are profound. Even when data is stripped of obvious identifiers, the conversational nature of Slack or email archives can inadvertently reveal personal details, project roles, or proprietary strategies. Experts such as Marc Rotenberg warn that re‑identification techniques could expose employees long after a company’s demise, potentially violating emerging data‑protection statutes like the EU’s GDPR or California’s CCPA. As AI developers scramble for richer datasets, regulators may need to clarify consent requirements for post‑mortem data sales, and firms could face litigation if anonymization proves insufficient.

For the AI industry, access to authentic workplace dialogues accelerates the training of autonomous agents capable of navigating corporate environments—scheduling meetings, drafting responses, or troubleshooting technical issues. This influx of high‑fidelity data could shorten development cycles and improve model reliability, giving early adopters a competitive edge. Yet the trade‑off between innovation and privacy may reshape market dynamics, prompting AI labs to invest in synthetic data generation or secure data‑sharing frameworks. Stakeholders across the ecosystem must balance the financial allure of these transactions against the ethical responsibility to protect employee privacy.