Spirit Airlines couldn’t survive the airline business, but its data might help build the next generation of AI products. Google has agreed to pay $10 million for internal business data from the bankrupt carrier, with plans to use it for product development and AI model training. It’s a strange transaction, but it tells you something real about where the AI industry is right now.
The data package includes employee emails, Microsoft Teams messages, spreadsheets, calendars, and marketing and operations data. Google says the data will be de-identified before the deal closes, meaning no customer information or personally identifiable data will be included. A U.S. bankruptcy judge is set to consider approving the sale at a court hearing this week.
Spirit, which shut down in May after collapsing under high debt and fuel costs, is liquidating whatever assets it has left. Internal communications and productivity data apparently count as assets now. And Google isn’t the only buyer interested. Mercor, an AI data company, submitted a competing bid of $7.5 million. The fact that two AI-focused buyers showed up for a bankrupt airline’s internal files says a lot about how tight the market for quality training data has become.
That’s the real story here. AI companies are running into the limits of publicly available data. The internet has been scraped. Common Crawl, Wikipedia, books, code repositories, all of it has been processed many times over by the major labs. So companies are getting more creative about sourcing new material. Corporate communications, spreadsheets, and operational data represent a category of text that reflects how people actually write and work, not how they perform for an audience online. For training models that need to handle business tasks, that kind of data is genuinely useful.
Google’s competitors are doing the same. OpenAI has pursued data licensing deals with publishers and companies. Anthropic has been similarly aggressive about data sourcing. The difference here is that bankruptcy proceedings create a rare opening to acquire data that would otherwise never be for sale.
There are real questions worth watching as this deal moves through the courts:
- Whether employees were adequately notified their communications could be sold
- Whether de-identification is sufficient given the sensitivity of internal email and chat data
- Whether bankruptcy law is the right mechanism for approving data transfers of this kind
- How regulators might respond as this type of acquisition becomes more common
$10 million is a rounding error for Google. But the precedent is what matters. If this sale gets approved without significant pushback, expect more AI companies to monitor bankruptcy filings the same way private equity firms do. Distressed corporate data could become a regular part of the training pipeline. That’s worth paying attention to.
