Google’s $10 million purchase of Spirit Airlines business records for AI training has turned a bankruptcy auction into a fresh test of whether sensitive workplace data can truly be made anonymous.

Google won the auction for a broad cache of records from the defunct airline, including calendars, documents, spreadsheets, emails and employee chats. The deal places routine internal communications—often rich with personal and operational detail—inside the expanding market for data that can sharpen AI systems.¹

The central dispute is not over Google’s stated purpose, but over the safeguards surrounding the material. The company says it will not take possession of raw personal data. “We will not receive any personal information from this dataset,” a Google spokesperson said, adding that an outside party will remove personally identifiable information before the files are delivered to Google.¹

Google further said any information it receives will be “rigorously scrubbed” by that third party.¹ That assurance is crucial because the purchased trove spans formats—emails, chats and calendars—that can reveal identities through context even when obvious identifiers are removed.

The sequence is straightforward: Spirit Airlines entered bankruptcy and its business records were put up for sale; Google then won the data package for $10 million; now the company’s de-identification commitment will define whether the transaction becomes a model for data reuse or another warning about the limits of anonymization in the AI race.