AI Companies Are “Harvesting Organs” from Dead Startups

In the race for ever-better AI, the obvious data sources — the entire public internet, books, Reddit, Wikipedia — ran dry by late 2024. Now the industry has found a macabre but lucrative new gold mine: the digital corpses of failed startups.

AI Companies Are “Harvesting Organs” from Dead Startups — And Founders Are Cashing In Slack archives. Jira tickets. Email threads. Google Drive folders full of internal docs. All the messy, real-world “operational exhaust” that once lived inside companies that no longer exist is being bought, anonymized, and fed into the next generation of agentic AI models.

The latest chapter in this quiet boom comes from a Forbes investigation published on April 16, 2026. When Shanna Johnson, CEO of video-captioning startup cielo24, decided to shut down her 13-year-old company, she turned to a specialized shutdown service called SimpleClosure. What she expected was help with payroll, taxes, and investor paperwork. What she got instead was an unexpected lifeline: an offer to sell her company’s entire internal workspace data to hungry AI labs.

Johnson walked away with hundreds of thousands of dollars — enough to pay off every remaining debt and give the team a clean break. “It’s cool to think that our data could be useful, live on and help other people,” she told Forbes, still emotional about closing the company.

From Shutdown Service to Data Broker

AI Companies Are “Harvesting Organs” from Dead Startups — And Founders Are Cashing In SimpleClosure, which helps founders wind down companies compliantly, has turned this into a full-blown business. In the past year alone it has completed nearly 100 such deals, recovering over $1 million for founders. Typical payouts range from $10,000 to $100,000 per company, though richer datasets (especially from older or larger startups in healthcare or finance) can command significantly more.

Demand is so intense that the company just launched Asset Hub — a dedicated marketplace (currently in beta for workspace data) where shutting-down startups can list their source code, Slack histories, Jira tickets, emails, and internal documents. The process is fast: after an initial assessment, a sale usually closes in 1–2 weeks. If no buyer appears immediately, SimpleClosure holds the data and sells it later.

CEO Dori Yona describes the frenzy bluntly: “There’s a feeling of a gold rush from these companies trying to get their hands on real-world data.”

Before any sale, SimpleClosure scrubs all personally identifiable information (PII). They call the process “rock solid” and are being extra careful while the workspace-data side of Asset Hub is still in beta. Competitors like Sunset are doing the exact same thing at similar price points.

Why AI Labs Want Your Old Slack Drama

AI Companies Are “Harvesting Organs” from Dead Startups — And Founders Are Cashing In Public datasets are too clean, too synthetic, too generic. What the new wave of “agentic” AI needs is the messy reality of actual workplaces: how people argue in threads, how tasks move through Jira, how decisions get made under pressure, how birthdays get planned in “Big Tech World.”

This data is now being poured into reinforcement learning gyms — simulated corporate environments where AI agents practice real tasks before being deployed.

Companies like AfterQuery have already built pre-packaged “worlds” (Finance World, Tax World, etc.).

Anthropic is reportedly considering spending $1 billion on such gyms, while startups like Prime Intellect (valued north of $1 billion) and Fleet are racing to build their own.

The Privacy Question No One Wants to Answer

Not everyone is celebrating. Privacy advocates point out that even after anonymization, career-spanning Slack conversations and emails contain deeply personal context. Marc Rotenberg of the Center for AI and Digital Policy told Forbes it’s “not generic data. It’s identifiable people.” Models have been shown to memorize and regurgitate training data verbatim in the past. If the scrubbing isn’t perfect, sensitive details could leak into future AI outputs.

For now, the gold rush continues. Every time a startup dies, its digital remains get a second life — training the very AI systems that may one day make even more startups obsolete.

Founders get a surprise payday. AI labs get premium real-world data. And the cycle of creative destruction in Silicon Valley just found a brand-new, slightly macabre revenue stream.