Google Buys Spirit Airlines' Entire Data Trove for $10 Million
Google won a bankruptcy auction for the internal data of defunct Spirit Airlines, paying $10 million for a trove that reportedly includes roughly 100 million emails and 500 million Microsoft Teams chats, to help train its AI systems.
What Google actually bought
Google beat out AI data company Mercor, which had bid $7.5 million, to win the auction. The package includes internal emails and Teams collaboration records, along with pricing data, flight operations and maintenance records, disruption and delay history, HR information, financial databases, audits, and marketing materials, in effect a full internal record of how an airline communicated, priced seats, managed staff and aircraft, and ultimately failed. According to a court filing, the data does not include personally identifiable information: it will be deidentified before transfer, and Google has agreed not to attempt to re-identify any individuals. The sale still required court approval at a hearing scheduled for August 19.
Spirit Airlines, which filed for bankruptcy and ceased operations, joins a growing list of failed companies whose internal data has found a second life as AI training material, a use nobody was thinking about when those emails and chat messages were originally written.
Why this matters to small and medium businesses
Read the data provisions in your vendor and partner contracts, especially what happens if that company goes bankrupt or gets acquired. This deal is a vivid, concrete example of something easy to overlook: your internal communications with a vendor, and anything you've shared with them, can become a saleable asset in a bankruptcy proceeding, potentially including as AI training data.
If your own business is in financial distress, don't assume your internal data is worthless or a formality to dispose of. Spirit's operational, HR, and communications data drew competing bids specifically because of its AI training value; get legal advice before assuming how that kind of data will or should be handled in a wind-down.
"Deidentified" is doing a lot of work in this story, and it's worth understanding what that promise actually covers. Before assuming your own customers' data would be handled the same way in a similar scenario, ask any vendor holding your data what deidentification actually means in their contracts, and who would decide how it's used if their business changed hands.
Sources: