Google has won an auction to buy the internal data of Spirit Airlines in bankruptcy court for $10 million, according to a notice from the U.S. Bankruptcy Court in the Southern District of New York.
The search giant reportedly plans to use it to train its AI models as public data sources are being exhausted.
The purchase includes about 100 million emails, 500 million Microsoft Teams messages and 30 million lines of software code, along with data from aircraft operations, pricing, marketing, human resources, finance and customer-service workflows, according to published reports.
Spirit’s records also include the pricing dataset of 7.2 billion competing flights and the transaction data of 7.5 billion passenger transactions, as well as information on operational records in marketing, HR, strategy and project management, according to Bloomberg Law.
The data to be handed over will be “de-identified,” meaning that it shall not include personal and nonpublic information of customers, according to the court. It would exclude any communication, documents or information. The transaction excludes major categories of customer information, including passenger profiles, loyalty-program records and credit-card information.
But the bankruptcy court delayed approval after the flight attendants’ union objected, saying employee data must also be protected alongside customer information, according to The Wall Street Journal.
The Association of Flight Attendants-CWA want all confidential flight attendant data to be excluded from the sale or at least have the same protections as consumer data, the paper said.
As AI companies seek increasingly specialized data to improve AI systems, the internal records accumulated by ordinary businesses are becoming valuable training assets. That is fundamentally different from much of the publicly available text used to train large language models.
An airline’s internal data can capture the messy interactions involved in operating a real business – how employees communicate, schedules change, prices are set, software is developed, customer problems are handled and operational decisions are made.
As AI expands beyond answering questions and generating content toward systems expected to perform complex workplace tasks, this enterprise data becomes increasing crucial. Models intended to operate inside companies may benefit from data showing how companies actually operate rather than relying primarily on information available on the public internet.
Notably, the second winning bid also came from an AI company: Mercor, which offered to pay $7.5 million.