Google Buys Spirit Airlines’ Emails, Files And Flight Records For $10 Million — So Gemini Can Learn From Bankruptcy

Google won a bankruptcy auction for Spirit Airlines internal data. They’ll pay $10 million for the failed carrier’s full messages, documents, and operating history. They beat out AI data company Mercor, which bid $7.5 million.<br>The customer list isn’t included in Google’s $10 million deal. Even in liquidation, Spirit still charges extra for the passenger.

The data will be used to train Google’s AI, and it’s almost too easy to mock. Spirit went bankrupt twice. At last, they can delay your answer six hours, lose the prompt and charge $39 to finish the answer. This likely solves the AI alignment problem, though, because training on Spirit Airlines data will never help Gemini take over the world.

In fact, Google isn’t buying Spirit’s fee-based business model, passenger brawls, or customer service. They’re buying a huge record of how a company communicated, priced inventory, managed employees and aircraft, responded when operations broke, and ultimately failed.

What Google Is Buying Out Of Spirit Airlines Bankruptcy

Google won the bankruptcy auction but it’s still subject to court approval at a hearing scheduled for August 19. Some of the employee data reaches back to August 1986. Most of the valuable operating data begins in 2008 and many systems cover only the last several years. It includes:

100 million emails, 500 million Microsoft Teams records, 17.1 million OneDrive files, 20.6 million SharePoint files and 667,563 IT tickets<br>516 source code repositories containing about 30 million lines of code, 372,585 commits and 43,170 pull requests, together with bug reports, review discussions, tests and build logs<br>Records covering 763,391 Spirit flights, more than 5 million crew pairings, aircraft maintenance and fuel, and roughly 3 billion irregular operations and passenger reaccommodation rows<br>190.3 million reservation records and 7.5 billion transaction rows, plus refunds, vouchers and flight credits (but personal information gets removed)<br>3.5 billion booking curve observations and 7.25 billion competitor flight pricing observations<br>Finance and accounting systems, forecasts, board presentations, audit and fraud workflows, deal pipelines, lender materials and M&A analyses<br>Human resources, payroll, training and employee workflow records (subject to deidentification)<br>Spirit-owned legal forms, policies, nonprivileged case files and contract histories showing drafts, redlines and final documents

This connects communication, decisions, software changes and operating results across an enterprise. Spirit’s archive will show an IT ticket, the emails about it, the code change, the review comments, the deployment result and what happened to the business afterward.

Why Google Wants Data From An Airline That Failed

Google’s explanation is that the Spirit dataset “can be helpful in improving our products and AI models.” Much of this value isn’t web-text pretraining. It is useful for training and evaluating AI agents that need to complete multi-step business tasks from read an email, to finding a policy, changing a forecast, writing or repairing code, routing for approval and explaining the result.

The code archive contains the full sequence around the 30 million finished lines. That means discussions, commits, tests and deployment logs that can teach a coding agent what people tried, what reviewers rejected and whether a change worked.

During a flight disruption, the carrier had to reconcile aircraft location, maintenance status, crew legality, gates, passenger connections, weather, inventory and cost. Spirit produced billions of records around pricing and reaccommodation which will support simulations and decision-making models where an action has a measurable consequence.

And by the way, Google Cloud just announced a five-year data and AI partnership with Ryanair covering productivity, customer service, fleet operations and maintenance scheduling. While it’s not likely Spirit’s data will be used directly for Ryanair, Google even has airline-specific reasons to want the operating knowledge.

The fact that Spirit failed spectacularly multiple times even helps as Google’s model learns what not to do, how problems compound and which warning signs appeared before an outcome. Spirit lost its cost advantage, could not sell enough premium product, suffered from a toxic brand, was unable to complete a merger due to antitrust and eventually ran out of financing.

The Limits Placed On Google’s Purchase

Google’s data acquisition doesn’t give it the full raw archive. It will have to spend heavily to meet several conditions:

Personal data must be removed or transformed by a third-party deidentification agent before Google receives it.<br>The agent must certify deidentification using the California Consumer Privacy Act standard for U.S. consumer data even where that law would not otherwise apply and HIPAA standards...