Google has acquired the massive corporate data trove of collapsed low-cost carrier Spirit Airlines for $10 million at a bankruptcy auction, in the latest sign that AI companies will pay real money for text that has never touched the open internet.
According to a court document filed last week and reported by The Register on August 18, 2026, Google's winning bid beat out at least one other serious contender: Mercor, a company that supplies data for training AI models. Axios separately confirmed that Google won the bankruptcy auction for Spirit's emails, chats and documents. For more context on this story, see our ongoing breaking AI news.
Spirit, once the archetypal ultra-low-cost US airline, never recovered financially from the shock that COVID-19 delivered to travel in 2020. After years of mounting losses, the carrier permanently grounded itself in May 2026 and entered liquidation, with its assets now being auctioned off to repay creditors. One of those assets, it turns out, was an archive of operational exhaust: emails, call recordings, chat logs and flight records — exactly the kind of vast, messy, human-generated text that AI developers say they need.
What $10 Million Bought
The numbers in the court filing describe a dataset of remarkable breadth. The trove includes more than 100 million emails and 500 million items from Microsoft Teams, alongside 17 million OneDrive files and 20.5 million SharePoint items. Google also acquired over 30 million recorded customer service calls and more than 15 million customer service chat records — a goldmine of real human dialogue, from booking changes to baggage complaints.
The archive reaches beyond customer interactions into the airline's operational core. The sale covers roughly 600,000 ServiceNow tickets, 13.7 million active email addresses from Oracle's Responsys marketing platform, and records of 11 million in-flight Wi-Fi sales. On the operations side, the data describes more than 763,000 flights, five million crew pairings, over 1.2 million fuel slips, and the purchase history of 787,452 aircraft parts.
Google has reportedly said it bought the data to improve its AI services. The Register noted that the search giant may see in Spirit's records the raw material for a domain-specific model — an aviation operations assistant, for instance — or simply a large corpus of everyday business records that could help models handle routine workplace tasks more credibly.
Why an Airline's Archives Are Suddenly Valuable
The auction illustrates a quiet shift in how AI companies source training data. The public internet has already been scraped repeatedly, and frontier labs now hunt for text that is fresh, specialized and — crucially — guaranteed to be human-written. Anything produced before 2022 is, by definition, free of AI-generated content, which matters because researchers have warned that models trained heavily on synthetic text risk "model collapse," a progressive degradation in output quality.
That same logic is driving stranger acquisitions. This week, TechCrunch and 404 Media reported that Amazon has been buying rare and out-of-print books, cutting off their spines and scanning them at a facility in Las Vegas — texts prized partly because they exist nowhere on the internet. Spirit's archive is the corporate equivalent: proprietary, unreachable by web crawlers, and dense with authentic exchanges between real customers and real support agents.
The presence of Mercor as the underbidder is telling. A company whose entire business is supplying AI training data was willing to come close to Google's price for an airline's emails and call logs — evidence that bankruptcy auctions have become a functioning marketplace for training corpora, with specialist brokers bidding against the giants they supply.
Privacy Questions Trail the Deal
Spirit's data was deidentified before being put up for sale, according to the court filing, and Google has promised to scrub any remaining personal information it finds in the archive.
Whether that scrubbing proves perfect is another matter. The Register pointedly noted that it awaits evidence of the "inevitable SNAFUs" — the near-certainty, in its telling, that some personal detail from a heated customer service call will eventually resurface in a model output or search result. Recorded phone calls and support chats are notoriously hard to fully anonymize: voices, names, booking references and itinerary details can all survive deidentification pipelines.
For former Spirit customers, there is no opt-out. Once data becomes a bankruptcy asset, it is sold like any other property the estate holds. Privacy advocates have long warned that liquidation converts personal histories — a distress call about a cancelled flight, a complaint filed after a family emergency — into inputs for products built by companies the customer never chose to share information with.
Data as the New Distressed Asset
The Spirit sale may foreshadow a broader pattern: distressed companies across retail, travel, telecom and finance are sitting on archives that AI developers may value more than physical equipment. For creditors, selling data alongside gates, brands and aircraft leases recovers extra cash. For AI labs, proprietary data is one of the few remaining ways to differentiate models trained on essentially the same public internet as everyone else's.
For Google specifically, $10 million is pocket change against its AI infrastructure spending. The strategic value lies in exclusivity: no rival can license Spirit's customer service dialogue after Google has bought it outright.
The deal also lands as regulators scrutinize how AI companies obtain training data. Lawsuits from publishers and authors over scraped content are working through US courts, and the EU's AI Act imposes transparency requirements around training data. Buying data outright at auction, with deidentification and PII-scrubbing commitments attached, is emerging as the compliance-friendly path — even if that privacy fine print will be tested the first time a Spirit customer's details surface in a chatbot's answer.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
