Why Google is paying $10 million for a failed airline's data
August 26, 2026
Based on reporting from Time Magazine → — simplified & explained by VAIIYA.
Spirit Airlines halted flights in May, but the airline's digital archive turns out to be a goldmine for the AI industry. Google won a bankruptcy auction this month with a $10 million bid for the defunct airline's internal company data. The tech giant's bid beat out competitor Mercor's ($7.5 million).
The acquired dataset includes roughly 100 million emails, 500 million Microsoft Teams messages, and about 30 million lines of code. According to the parties involved, customer data is not part of the sale.
From software code to regular office work
AI developers have made major progress recently in building so-called coding agents — AI assistants that write software on their own. Programming lends itself well to this: written code either works or it doesn't, which provides a direct, objective feedback loop for the model.
To deploy AI models for regular office work outside the IT sector, however, requires a completely different kind of data. Emails and chat messages show how people actually communicate, collaborate, and make decisions within a complex organization.
Reinforcement learning and realistic data
The purchase is closely tied to the rise of reinforcement learning (RL). In this training method, AI agents perform tasks in simulated software environments and are rewarded when an action leads to the desired outcome. AI lab Anthropic has previously said it's setting aside more than a billion dollars a year to develop these kinds of environments.
According to Nick Heiner, head of reinforcement learning at data company Surge, synthesized or human-generated test data isn't always sufficient. Real data from an operating company shows how human employees actually function. Even inefficient processes or bad decisions in Spirit Airlines' archive can be valuable for teaching models how workflows could be improved.
Privacy and legal objections
The transaction isn't without controversy. Although a third party removes all personal information before Google receives the dataset, a flight attendants' union has formally objected in court. The union fears that because the relationships within the dataset are preserved — so-called referential integrity — anonymized information about employees could still be traced back to them. The judge is expected to rule on the case on September 9.
Meanwhile, another AI data specialist, Micro1, submitted a higher bid of $12.5 million after the auction had already closed. Regardless of which party ultimately secures the dataset, the fight over the Spirit data makes clear that the AI industry has set its sights on automating broader office work.