Opinion: Spirit's Workers Lost Their Jobs. Now Their Emails Are Worth $10 Million.

When Spirit Airlines shut down in May, its planes went to whoever would take them, its gates went back to the airports, and its workers went home with final paychecks. But the airline left something behind that turned out to be worth more than anyone expected: its own memory. In August, Google won an auction for the digital record of how Spirit operated, and on October 8 more than 120 members of Congress demanded the deal be stopped.

The price was $10 million. The inventory, according to reporting on the auction, includes roughly 100 million employee emails, 500 million Microsoft Teams messages, 17 million OneDrive files, 20 million SharePoint items, and 516 code repositories holding 30 million lines of custom software. The personal information of 97 million customers has been carved out at a privacy ombudsman's recommendation, and a bankruptcy judge is expected to rule on the sale at an October 14 hearing. The employee data, however, is still on the block.

Think about that for a moment. Not the ombudsman, not the $10 million. Think about the emails. Millions of messages written by people who were doing their jobs, complaining about schedules, joking with colleagues, forwarding a medical accommodation request, who never once imagined their words would be sold at auction to train an artificial intelligence.

What Google is really buying

Google says it does not want personal information, and that any such data will be excluded or de-identified by an independent third party before it arrives. Take that at face value. Even stripped of names, the corpus is worth a fortune, because what AI companies are starving for now is not more scraped web pages. It is real workplace data: the emails, chats, spreadsheets, and code that show how actual humans do actual jobs.

Real-world data like this is what trains AI systems to handle complex business tasks, which is exactly the market every major lab is chasing. A model trained on how an airline really operated is a model that can draft the email, reconcile the spreadsheet, and debug the code. Spirit's employees spent years doing that work for wages. Google is about to buy the residue of it for $10 million. The people who made the data do not get a cent.

Your work emails were never yours

Legally, this is all perfectly ordinary. Under standard employment arrangements, work product belongs to the employer, and the inbox you used every morning belongs to the company, not to you. When a company liquidates, its assets get sold: the leases, the trademarks, the servers. Now, apparently, the emails too.

But legal is not the same as legitimate. The law assumed that a company's records would pass to a successor who would run the business, not to a technology giant training a machine to automate the next hundred thousand jobs like yours. Bankruptcy was designed to wind down what a company owed. It was never meant to be a loophole that converts workers' private correspondence into AI training data without their knowledge, let alone their consent.

The law assumed a company's records would pass to a successor who runs the business. It did not imagine them auctioned off to train a machine.

De-identification is a fig leaf

Notice the asymmetry in how the deal is being handled. The ombudsman recommends excluding passenger data outright, because passengers are customers and customers get privacy. Employees are the product here, and their data gets a weaker promise: scrub it, anonymize it, and move along.

De-identification sounds reassuring until you think about what it means applied to 100 million emails from one company. Strip the names, and you still have the scheduling disputes, the medical accommodation requests, the union organizing threads, the manager who documents every grievance in writing. This is the kind of corpus where a pattern is a fingerprint. Anyone who has worked in privacy knows that anonymizing dense interpersonal records is genuinely hard. That difficulty is exactly why the passengers' data is being excluded rather than scrubbed.

The lawmakers' letter, led by Senator Elizabeth Warren and Representative Steven Horsford, asked for employee information to be excluded "to the greatest extent possible," and for no transfer to happen until real protections are in place. Their core claim is hard to argue with: innovation should not come at the expense of workers' privacy, or of information people were required to provide as a condition of employment.

The precedent this sets

Here is why this matters beyond one dead airline. Every bankrupt company in America has a data room: the emails, chats, and documents of its former workers, sitting on a server somewhere. Until this year, that data was a liability, something to be wiped or archived. Now it is an asset with a price tag. Spirit's auction has put a number on it, $10 million for a midsize airline, and every future liquidator is taking notes.

The market logic is obvious and unsettling. If workers' communications are sellable assets, then every message sent on a company account becomes inventory, waiting for the company's balance sheet to fail. The workers never signed away their consent because they never had it to give; the data was already gone the moment they pressed send. Microsoft president Brad Smith recently argued that AI has fewer safety rules than milk, and on this evidence he has a point. Dairy gets more regulatory scrutiny than the wholesale transfer of a workforce's words.

There is a better answer, and it is not complicated: the people who produced the data should get a say in its sale. Congress could extend to employees the same protections the passengers get, making employee data off-limits in asset sales without affirmative consent, or requiring that proceeds flow back to the workers whose words created the value. Either would recognize a simple fact: a workforce's communications are the output of labor, and labor has rights that a bankruptcy auction should not extinguish.

What happens next

The judge's ruling, expected at the October 14 hearing, will decide this sale. But the longer game is the one the 120 lawmakers are really playing: establishing that this category of transaction, bulk employee data sold for AI training, cannot happen quietly, on the cheap, and without consent. The question hanging over the whole AI industry is who gets to profit when the raw material of training data is other people's lives.

Spirit's workers already lost their jobs. They should not lose their emails too. And if the industry insists that its future depends on data like this, then the people who wrote it deserve a seat at the table, and a share of the $10 million.