For 25 years, no major U.S. airline had simply switched off the lights and walked away. Spirit Airlines did it in May. Three months later, the most contested asset left in the wreckage is not a jet, a gate slot or a maintenance hangar — it is the company inbox.
Alphabet’s Google has agreed to pay $10 million for Spirit’s internal enterprise dataset, a haul that bankruptcy court filings put at roughly 100 million emails and 500 million Microsoft Teams messages, plus spreadsheets, calendars, marketing materials, HR records, financial databases, audits and project-management documents. Google won the material at a bankruptcy auction, outbidding the AI data startup Mercor, which offered $7.5 million and was named backup buyer should the deal collapse.
“We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models,” a Google spokesperson said in a statement. “We will not receive any personal information from this dataset. Any data we receive will be rigorously scrubbed of any personally identifiable information by a third party before receipt.”
The sale goes before Judge Sean Lane of the U.S. Bankruptcy Court for the Southern District of New York on Wednesday.
What is actually in the box
The raw archive is considerably larger — and more personal — than what Google says it will ultimately receive. Court documents describe nearly 100 million passenger names, close to 13 million active email addresses and nearly 176,000 employee records, some reaching back to 1986. Filings also reference roughly 7.2 billion records covering competitors’ flights and about 7.5 billion passenger transaction records dating to 2008: an unusually deep map of how one carrier priced seats against its rivals for nearly two decades.
Excluded from the transfer, according to the estate, are 97.5 million passenger profiles and 50.2 million Free Spirit loyalty records. Dylan Friesner, a vice president at PJT Partners and an investment banker for Spirit, told the court the transferred data contains no personally identifiable information. A court-appointed third-party ombudsman is to supervise a process the filing calls “deidentification,” defined as rendering data that “cannot be associated with, reasonably used to infer information about, or otherwise linked to, a particular consumer.” Google separately agrees not to attempt re-identification.
Spirit’s flight attendants are not reassured. “This is outrageous! We are filing a court objection to Google’s attempt to buy data that has no business being sold,” said Sara Nelson, international president of the Association of Flight Attendants-CWA, which still represents more than 5,500 of the airline’s former crew members. The union’s objection goes at the question the hearing has to resolve: whether workers who wrote hundreds of millions of internal messages under an employment relationship that no longer exists retain any say over where those messages end up.
Why it matters
The frontier labs have largely strip-mined the open web. What they still lack is the messy, sequential, consequential record of how work actually gets done — the thread where a scheduling conflict gets escalated three levels, the spreadsheet revised six times before a decision, the chat where someone says the plan will not work and explains why. That is precisely the training signal agentic AI is starved of, and precisely what a company’s internal comms archive is made of.
Bankruptcy is turning out to be the cleanest way to buy it. A living company will not sell its email archive; the reputational and legal exposure is too great. A dead one has a fiduciary duty to creditors to sell everything not nailed down. Forbes has reported that AI firms are increasingly acquiring defunct companies’ Slack archives, Jira tickets and Drive files for exactly this reason — Spirit is simply the largest and most visible example yet.
The price is the tell. Ten million dollars is a rounding error against Alphabet’s capital expenditure, which runs to tens of billions a year on AI infrastructure alone. Yet it buys a corpus no amount of synthetic data generation reproduces, because the value is in the friction: the wrong turns, the reversals, the institutional context that only accumulates when real people are on the hook for real outcomes. Box chief executive Aaron Levie, posting on X about the deal, put it bluntly: “When you hear that data is the new oil, this is ultimately what that looks like.”
The auction dynamic matters as much as the number. Mercor — a training-data company founded by some of the world’s youngest self-made billionaires — bid within 25% of Google. Two very different buyers converged on the same asset class, which suggests a market forming rather than a one-off curiosity. Expect corporate archives to start showing up on liquidation schedules as line items with estimated values attached.
What to watch
Wednesday’s ruling is the immediate event. Judge Lane must weigh the estate’s obligation to maximize creditor recovery against Nelson’s objection and any scrutiny from the U.S. Trustee, the Justice Department’s bankruptcy watchdog, which has not publicly commented. If he approves, the order becomes a template: a court-sanctioned procedure for converting a dead company’s internal communications into AI training material, complete with an ombudsman and a deidentification standard other estates can copy.
Watch, too, whether deidentification holds up in practice. Stripping names from 100 million emails is not the same as making them unlinkable; writing style, project references and org-chart context are all re-identifying signals, and the academic literature on deanonymization has been consistently unkind to “reasonable measures.” Any post-transfer audit of that scrubbing will be worth more than the press releases.
The broader question is whether Spirit is the last estate to do this quietly or the first to do it loudly. Every company that files Chapter 11 from here forward has a newly legible asset on its balance sheet — and every employee at those companies just learned that their inbox is on it.
“This is outrageous! We are filing a court objection to Google's attempt to buy data that has no business being sold.”— Sara Nelson, International President, Association of Flight Attendants-CWA