NewsStocksGoogle to Pay $10 Million for Spirit Airlines Data to Train Its AI Models

Google to Pay $10 Million for Spirit Airlines Data to Train Its AI Models

Author: Decrypt·

Key Takeaways

  • •Google’s $10 million agreement covers Spirit Airlines’ internal business data, which the company says will help improve its products and AI models.
  • •The dataset reportedly includes emails, spreadsheets, bookings, frequent-flyer information, Microsoft Teams messages, calendars, and other operational records.
  • •Google says the information will be de-identified before it is transferred and that no personal information will be received.
  • •Spirit halted operations in May and is selling remaining assets through bankruptcy, with the deal still awaiting court approval.
  • •AI data company Mercor reportedly submitted a competing $7.5 million bid for the same dataset.
Google to Pay $10 Million for Spirit Airlines Data to Train Its AI Models

Google has agreed to pay $10 million for Spirit Airlines' business data, which the company plans to use to improve its products and AI models.

The deal covers internal communications and company records, including emails, spreadsheets, bookings, frequent-flyer information, and employee human resources data, according to a report by CNN. The information will be stripped of details that could identify individuals before Google receives it.

"We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models," a Google spokesperson told CNN. The company said it will not receive personal information through the purchase.

In a separate report, Reuters said the data also includes Microsoft Teams messages and calendars, as well as marketing, productivity, and operations records.

Spirit halted operations in May and has been selling its remaining assets through bankruptcy. Airline bankruptcies typically end with another carrier buying the company, including its data. Selling data as a standalone asset in an insolvency has drawn regulatory scrutiny before: in 2000, the Federal Trade Commission stepped in to stop failed online retailer Toysmart from selling its customer list, and in 2011 Borders was permitted to transfer customer records to Barnes & Noble only after the FTC imposed conditions, including giving customers a chance to opt out. Google says the Spirit dataset is different because it will be de-identified before it changes hands.

Google faced competition for the dataset: AI data company Mercor reportedly submitted a $7.5 million bid. The sale still requires court approval, and U.S. Bankruptcy Judge Sean Lane is scheduled to consider the deal at a hearing Wednesday.

The sale comes as AI companies increasingly pay for access to data used to train and improve their models, reflecting demand for training material beyond what is freely available on the public web. In February 2024, Reddit gave Google access to nearly two decades of user-generated content in a deal reportedly worth $60 million a year. In May 2024, OpenAI reached its own agreement to bring Reddit content into ChatGPT. More recently, in January, the Wikimedia Foundation announced agreements with Microsoft, Google, Amazon, Meta, and other AI companies for large-scale access to Wikipedia content.

The Spirit transaction differs in kind from those arrangements: rather than licensing content from an operating platform, Google is buying the internal records of a company that no longer exists, and whether the court blesses the price and terms will be the next step to watch.