Google has won an auction for Spirit Airlines' data with a $10 million bid, according to a notice filed August 14 in the airline's Chapter 11 case in the Southern District of New York, which labels the trove "the Deidentified Data."
The auction only settled who bid highest, though, and nothing moves until the court approves the sale, and that hearing is today.
Spirit stopped flying on May 2 of this year, halfway through its second bankruptcy in two years, and the estate has been selling off anything with a price ever since. Its LaGuardia takeoff and landing slots went to JetBlue last month. The data was what remained. And now come the vultures.
Most of it is the ordinary residue of running an airline. Google's bid takes in 100 million emails, half a billion items from the company's Teams chats, the OneDrive and SharePoint files behind them, and about 30 million lines of Spirit's own source code across 516 repositories.
That includes the operational record: 763,391 flights flown, more than a million fuel slips, the crew pairings, the parts logged in and out of the hangar. The employee files reach back to 1986l payroll, tax forms, 175,000 personnel records, most of them belonging to people who no longer work there.
The passengers are in there too, whatever the deidentified label is meant to suggest. Filed under revenue are more than 190 million booking confirmations and over seven billion transactions, going back to 2008.
The most obviously personal material is the part being held back. The schedule marks 97.5 million customer profiles and 50 million Free Spirit accounts as excluded, and with them more than 30 million recorded service calls, the chat logs, and the 13.7 million addresses in Spirit's marketing email system.
The estate is also keeping one thing back for another sale: under the agreement, Spirit stays free to sell its "customer data list, which includes individual traveler spend aggregated by year," to buyers in the hospitality or travel industries.
Google said in a statement that it had "acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models," and that the material will be "rigorously scrubbed of any personally identifiable information by a third party before receipt." A Google spokesperson told Forbes that "we will not receive any personal information from this dataset."
The estate goes further still. In a court filing, Dylan Friesner, the PJT Partners vice president working as Spirit's investment banker, said the data holds no personally identifiable information at all.
The filing describes data that "cannot be associated with, reasonably used to infer information about, or otherwise linked to, a particular consumer." That promise has to hold up against Google's own products, which are built to do the very thing deidentification is supposed to prevent: find the connections across records like these.
The sale agreement makes that easier rather than harder. It requires the scrub to preserve referential integrity across the data set, the links between records are kept intact on purpose, so a single pseudonymous employee could still be traced from a mailbox to a payroll file to a crew pairing. And the agreement lets Google choose the firm that does the scrubbing, and pay for its work.
A court-appointed ombudsman will oversee the process, but the cleaning itself is carried out by a contractor the buyer selected, at the buyer's expense, to a standard the buyer is entitled to weigh in on.
The buyer may not even be Google. If this sale falls through, the backup bid of $7.5 million belongs to Mercor, a company whose business is supplying data to train AI models, so the trove is bound for a model either way. Mercor wanted it more than the price suggests: at one point it offered $7 million for the chance to take the raw data first and anonymize it itself.
Not everyone is treating the sale as routine. Sara Nelson, international president of the Association of Flight Attendants-CWA, said the union is filing a court objection to what she called an attempt to buy data that "has no business being sold." The roughly 17,000 people who lost their jobs when Spirit folded are in the trove as well; their old inboxes, their pay records, 1,092,000 time cards.
