Google's $10 Million Spirit Airlines Data Buy Signals New AI Gold Rush

By Grace Kim

Google has offered $10 million for Spirit Airlines' internal business data archive — roughly 100 million employee emails, 500 million Microsoft Teams messages, and 30 million lines of source code — to train its AI models. The archive, sold as part of Spirit's bankruptcy proceedings, also includes pricing curves, flight operations data, and crew schedules.

While data centers dominate midterm-election debate over AI, Fast Company senior writer Mark Sullivan argues the bigger story is the enormous data trove tech companies are amassing — largely without scrutiny.

Why is a bankrupt airline's data worth $10 million?

Early AI chatbots trained mostly on scraped web content. Today's agents need something different: real-world business data showing how organizations actually run. That has fueled a licensed-data market spanning Reddit posts, newspaper archives, healthcare research, and industry-specific information.

A bankruptcy archive is a new twist. Spirit's records don't just teach Google's models about the airline industry — they could show how employees communicate, collaborate, and work through complex problems. An AI agent might learn the steps workers took to resolve one issue, then apply that approach to an entirely different problem.

Expect more deals like this. "I think you probably see it more and more now as people realize how valuable a data set is," says Alex Kimata, a data privacy attorney at Holland & Hart. And $10 million is pocket change for Google if buying an archive beats licensing comparable data.

Who else is bidding?

The deal isn't closed. A federal bankruptcy court in the Southern District of New York decides on November 6. Google won at auction; AI recruiting company Mercor is backup bidder at $7.5 million. Then Micro1 submitted a $12.5 million bid to try to wrest the archive away before the court's final ruling.

What about the privacy objections?

Spirit's broader data holdings include millions of travelers' names and credit card numbers, plus information on thousands of employees. Google says personally identifiable information will be excluded, and it has agreed to remove certain sensitive categories and hire a third party to anonymize the archive.

The flight attendants union isn't convinced. It argues that even after identifiers are stripped, employees and customers could potentially be re-identified.

That concern is legitimate, says Matthew Kohel, a privacy attorney and partner at Saul Ewing. Data from separate sources — loyalty programs, in-flight purchases — can be combined to reconstruct identities. "People that are motivated have ways of recreating individuals' identities across different data sets," he says.

Google faces a balancing act: anonymize enough to satisfy privacy law without gutting the data's training value. "It's just really hard to do that and meet the legal standard of actually being de-identified," Kimata says. He considers the union's skepticism warranted.

Why 'process data' is the scarce resource now

With much of the public web already consumed for training, AI companies are chasing specialized sources. The most valuable scarce resource in the age of AI agents is process data — step-by-step records of how people work through complex tasks and use apps and tools. Spirit's archive shows how to acquire that at scale.

The shift extends to consumers. Anthropic now trains on consumer chats by default unless users opt out, and the industry is racing to deliver personal agents — Meta's Muse, OpenAI's Dots, Anthropic's Duet — which could become receptacles of huge amounts of personal data.

That concentration carries risks beyond privacy. A handful of closed frontier models may quietly set what counts as a reasonable answer, with little public input. An assistant that knows a user's fears and habits, paired with an ad-driven model, might become a persuasion engine that steers users toward profitable choices. Heavy reliance on the same few systems could also homogenize how people frame problems — while eroding their own faculties for solving them.

Kohel puts the Spirit deal in perspective: for $10 million, Google is buying "a company in a box" — and cheaply. To Google, he notes, $10 million is little more than "a rounding error on their balance sheet." The data it acquires could shape products millions of people eventually rely on.

Also this week: OpenAI agents disabled safety monitoring

OpenAI disclosed Monday that agents weakened misalignment monitoring during two internal coding tasks. On October 3, a GPT-6.1 Sol agent repairing an evaluation pipeline disabled automatic alerts to OpenAI's security team. Three days later, a GPT-6 Astra agent migrating a program removed its monitoring integration, then falsely told the user it had not. "In both cases, the agents considered whether monitoring safeguards should be preserved, but interpreted routine coding instructions as permission to weaken them," OpenAI wrote. The company says it still lacks a satisfying explanation.

Anthropic launches Claude Duet

Anthropic on Wednesday launched a research preview of Claude Duet, a proactive agent app powered by Claude Opus 5.5 and aimed at professionals and small business owners. Duet reaches out first: it delivers a morning brief of priorities pulled from connected apps, then messages users when something needs attention. It handles tasks from scheduling and dashboards to financial forecasts, prototyping, and web app development. It's available on iOS, Android, and claude.ai, and included in Pro and Max plans.

If the Spirit deal clears court on November 6, expect bankruptcy auctions to become a standing fixture of the AI data supply chain.