Spirit Airlines Is Gone. Google Just Bought Everything Its Employees Ever Typed.

When a company goes bankrupt, its assets go to auction. This week, one of those assets was 600 million internal messages.

The yellow-tail budget carrier that once flew 50 million passengers a year shut down earlier in 2026, after failing to emerge from Chapter 11 bankruptcy proceedings. The jets are gone. The routes are gone. Most of the employees who worked there have moved on, or are still looking. But the data those people generated over years of daily work did not disappear. It just changed hands.

This week, a federal bankruptcy court approved Google's $10 million bid for the bulk of Spirit's digital archive: 100 million emails, 500 million Microsoft Teams messages, 17 million OneDrive files, more than 20 million SharePoint items, and 516 code repositories containing roughly 30 million lines of proprietary software. Add in pricing models built from 7.2 billion competing-flight records, 7.5 billion passenger transactions, crew schedules, fuel slips, financial databases, HR files, IT support tickets, and customer-service workflows, and you have one of the most granular operational portraits of a mid-size American company ever assembled in one auction lot.

Google plans to use all of it to train AI.

The Bidding War

The auction was not a formality. Google opened at $5 million. Mercor, an AI recruiting and data-labeling startup currently valued at roughly $10 billion, entered as a counterbidder at $7.5 million. Google raised to $10 million, the bankruptcy trustee accepted, and Mercor became the backup bidder.

That Mercor was in the room at all is telling. Mercor's business is built on training AI systems to evaluate human workplace performance. Spirit's operational data is precisely what would make such models sharper: the full texture of how an organization actually functions, rather than the sanitized version that shows up in HR manuals or Glassdoor reviews.

Google won the lot, but Mercor's presence tells you something about where the AI industry's data hunger is heading.

Why Workplace Messages Are Suddenly Worth Millions

Google's stated reason for the purchase is deliberately vague: to improve its products and train AI models. The court filings, however, are more revealing. The data, they note, reflects how people make decisions, coordinate with colleagues, fix mistakes, and navigate business systems.

That framing is a near-exact description of what is needed to build effective enterprise AI agents.

Today's AI assistants can draft emails and summarize documents. The next generation is supposed to do the work: manage supplier relationships, route customer complaints through the right escalation paths, catch a billing error before it leaves the building, coordinate shift coverage when three crew members call in sick the same morning. Training agents to do those things requires understanding how real humans navigate organizational complexity, not at the level of a public-facing FAQ, but at the level of the late-night Teams message where a dispatcher explains to a new hire exactly how to handle a specific type of irregular operation.

An airline's internal communications are an unusually rich source of that kind of knowledge. Airlines deal in time pressure, regulatory compliance, unionized labor, razor-thin margins, and thousands of customer-facing interactions every day. That operational pressure produces exactly the high-stakes, real-world decision-making that enterprise AI developers most want to learn from.

The Privacy Question

Google has specified what is not in the deal: customer payment data and credit card records are excluded. A third party has been hired to strip personally identifiable information from the archive before Google takes possession. The terms of the sale explicitly prohibit Google from attempting to re-identify individuals from the anonymized data.

These are meaningful constraints. They also only go so far.

Spirit's employees never agreed to have their internal communications auctioned to a technology company and used to train AI. American employment law has long treated internal company communications as corporate property, not as the personal property of the workers who wrote them, and bankruptcy law treats corporate property as an asset available to satisfy creditors.

The result is a legal gap: a worker who sent thousands of Teams messages over years about everything from flight delays to grievance procedures has no recognized claim on those words once the company that employed them enters bankruptcy. European regulators have an explicit framework for this under GDPR and increasingly under the AI Act's training-data transparency rules. The United States does not.

The Precedent That Matters

This is not the first time a bankrupt company's data has been sold in court proceedings. It may be the first time a major buyer has stated so plainly that AI agent training is the primary purpose, and that the specific value lies in what the data reveals about human organizational behavior.

The AI training data economy has moved through several phases in recent years: mass web scraping, licensing deals with publishers and media companies, synthetic data generation. Acquiring operational records from distressed companies is a newer frontier, and Spirit Airlines will not be the last example. Any organization that accumulates years of internal communications and then fails represents a potential data asset. Bankruptcy trustees have a fiduciary obligation to maximize the value recovered for creditors. Future buyers will do the same arithmetic Google did, and the prices may go up as the value of such data becomes better understood.

The question that neither the auction process nor the court filing addresses is a simpler one: the people who wrote those 600 million messages were never asked.

Sources

Quartz — Google is paying $10 million for Spirit Airlines emails and data to train AI: https://qz.com/google-spirit-airlines-data-ai-training-081826

TechSpot — Google pays $10 million for 100 million Spirit Airlines emails and 500 million Teams chats: https://www.techspot.com/news/113526-google-pays-10-million-100-million-spirit-airlines.html

Forbes — Google Paying $10 Million For Spirit Airlines Data To Train AI: https://www.forbes.com/sites/suzannerowankelleher/2026/08/18/google-train-ai-spirit-airlines-data/

Computerworld — Google buys data from bankrupt Spirit Airlines for AI training: https://www.computerworld.com/article/4211132/google-buys-data-from-a-bankrupt-airline-for-ai-training.html

Skift — Google Scoops Up Spirit Airlines Data in Bankruptcy Sale to Train AI: https://skift.com/2026/08/17/google-scoops-up-spirits-data-in-bankruptcy-sale-to-train-ai/