Google bought a dead airline's data. Check your contracts.
Google paid $10M for Spirit Airlines' data in a bankruptcy sale to train AI. A vendor bankruptcy can put your operational data on the auction block too.
A bankrupt airline's inbox is now AI training data. Google won a bankruptcy auction for roughly $10 million to acquire part of Spirit Airlines' enterprise dataset and source code, and the deal is stuck in court because two parties say some of what is being sold was never Spirit's to sell. A vendor bankruptcy can put your operational data on the auction block, and almost nobody reads their contracts for that scenario until it happens.
What actually happened
Spirit stopped flying in May. Computerworld reports the package Google bought includes roughly 100 million emails, 500 million messages including Microsoft Teams data, revenue and flight operations records, marketing, personnel and project management data, about 30 million lines of program code, pricing data on 7.2 billion competing flights, and around 7.5 billion passenger transactions. Google says a third party will strip personally identifiable information before transfer, and that the material will be used to improve its products and AI models.
Two objections have landed in the US Bankruptcy Court for the Southern District of New York. The Association of Flight Attendants-CWA objected in August, arguing the privacy design of the deal is aimed at consumers while the payload is disproportionately employee data: "The privacy architecture of this transaction is consumer-facing; its payload is disproportionately employee-facing." The union warned that de-identification preserving referential integrity across the dataset could still let records be traced back to individuals. It did not try to block the sale, only to get employee-identifying information removed.
The second objection is the one every business owner should read twice. Springshot, a San Francisco software vendor that ran technology operations for Spirit starting in 2022, filed a limited objection asserting that its 2022 software agreement reserves ownership of its software and of data generated by it — and that the sale sweeps up IP that belongs to Springshot, potentially handing Google material to build a competing product. The hearing that was expected to approve the deal was pushed back.
Why data ownership matters for your business
You are not in Chapter 11. Your vendors might be. Every SaaS tool you run — your POS, your helpdesk, your scheduling app, your 3PL portal — holds a working record of how your business operates. In a liquidation, that record is an asset, and the estate has a duty to sell assets for the highest price. AI labs are currently the highest price.
Three things to do this quarter, and none of them require a lawyer to start:
Find the assignment and bankruptcy clauses. Most contracts say the agreement can be assigned in a merger, acquisition, or sale of substantially all assets. That sentence is how your data ends up with a buyer you never chose. Note which vendors have it and which do not.
Find out who owns derived data. Your raw records are usually yours. The interesting question is who owns the data the platform generates about your operations — the models, the aggregates, the benchmarks. The Springshot dispute is exactly this fight, and it is being litigated because the contract language was clear enough to argue over. Yours probably is not.
Get your export running before you need it. A contractual right to export is worth nothing if nobody has ever tested it. Pull a full export from each critical vendor this month, confirm it is in a format you can actually load somewhere else, and put that on a schedule. A vendor in distress is not going to prioritize your data request.
The cheap version of this work is a spreadsheet: vendor, what they hold, assignment clause yes/no, derived-data ownership, last successful export. That document is worth more than most of the AI tooling you are being pitched.
Key takeaways
- Google won a bankruptcy auction at about $10M for Spirit Airlines' enterprise data and source code, to train AI models
- The package includes roughly 100M emails, 500M messages, 30M lines of code, and about 7.5B passenger transactions
- The AFA-CWA objected that the privacy design targets consumer data while the payload is largely employee data
- Vendor Springshot objected that its 2022 agreement reserves ownership of its software and the data it generates
- Assignment and bankruptcy clauses in your own SaaS contracts decide whether your operational data is sellable
- Test your data export from every critical vendor now — a right you have never exercised is not a right
If your main software vendor filed tomorrow, could you get your data out — and would you know who owns it? We build systems on stacks you own, with exports that run on a schedule instead of on a promise, so a vendor's balance sheet is never your operational risk. See how we scope vendor-agnostic builds, or send us your vendor list and we'll flag the lock-in.
Sources: Computerworld, Fortune, Law360.
- #data-ownership
- #vendor-risk
- #contracts
- #ai-training
- #procurement
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Senate AI duty of care draft: model releases get a veto
Thune, Cruz and Klobuchar are drafting a federal AI duty of care with power to block unsafe model releases and preempt state law. What it means for your stack.
Read itOpenAI won't IPO in 2026: your core vendor stays private
Sam Altman says an OpenAI IPO in 2026 would be 'ill-advised' given safety concerns. What a private core AI vendor means for the small businesses built on its API.
Read it