Who Owns AI Training Data? | Elacity
A court is weighing whether a bankrupt airline's internal data goes to Google to train AI. It sold for 10 million dollars, and the people who made it own none of it.
A Bankrupt Airline's Data Sold for 10 Million. Who Owns the Data AI Trains On?
Spirit Airlines stopped flying in May. Its data did not. This week a court in New York is weighing the sale of that data, roughly 100 million internal emails and half a billion workplace chats, to Google to train AI models. The people who wrote every one of those messages are not party to the deal, and they will not see a cent of the price.
The price is the part worth pausing on. Google won the auction with a 10 million dollar bid, beating a 7.5 million dollar offer from an AI data firm that was named the backup buyer. A defunct airline's operational exhaust, its pricing models, booking curves, and refund histories, just cleared a market. Data like this used to be a byproduct. It traded like an asset.
So the honest question under the headline is who owns the data AI trains on. Not who can access it, not who deleted it, but who holds equity in it once it turns out to be worth something. Spirit's answer, arrived at by bankruptcy, is nobody who made it.
The Sale Is Legitimate. That Is the Uncomfortable Part.
This is not a breach or a data grab. The sale is court supervised, the trove is being deidentified with the buyer barred from re-identifying anyone, and it explicitly excludes data on Spirit's 97.5 million passengers and 52.4 million loyalty members. By the rules we have, it is clean.
The discomfort is structural, not legal. A functioning market decided a dead company's internal record was worth eight figures to train a frontier model. The event that surfaced that value was liquidation. It took the company failing for anyone to price what it knew.
The Value Outlived Everyone Who Made It
The trove reportedly includes roughly 3.4 million payroll records and employee data going back to 1986. Four decades of people did the work that produced the pricing intelligence and the operational patterns now sold as training material. None of them held a share in it. Their output was booked as a cost of employment and is being sold as capital.
This is the shape of the whole data economy, drawn unusually clearly by a bankruptcy docket. You generate value continuously. Someone else records it as an asset. The gap between those two facts is the Sovereign Capital question: who holds the productive property. Right now, in that gap, nothing lives.
The popular fix is a data dividend, a small recurring check for the raw material you emit. We have argued that a dividend is the wrong shape for the problem, because a check is income and income can be cut. Equity is ownership, and ownership is what survives the counterparty. The Spirit docket is a reminder that data is already capital. The only open question is whose.
Ownership Has to Be Built In Before the Auction
You cannot retrofit ownership onto 100 million emails after the fact. It has to be a property of the data from the moment it exists. That is the specific thing Elacity builds, and it is worth naming the mechanism rather than waving at ownership.
With Elacity dDRM, a dataset is sealed into an encrypted, programmable good, a Wealth Capsule, with its rights and terms written in. An AI buyer can pay to use it, but the file is never handed over in the clear. The key that unlocks it is used, never owned: it exists in readable form only for a split second inside a sealed sandbox, welded to one authorized use, then wiped. No buyer, platform, or attacker ever holds it.
That is the difference from an ordinary data sale. Today's data licensing marketplaces will pay you and still take the file; once the copy leaves, your terms are a promise. Elacity enforces the terms at the key layer instead, where independent machines re-check your on-chain rights before releasing their share of the key. Ownership stops being a clause you hope holds and becomes a condition of access.
Be precise about what runs. The hard primitive, a key an AI can use but never see, already works today. The consumer on-ramp, where anyone wraps their own data into a Wealth Capsule and sets its terms without touching cryptography, is still being built. Naming that line is the point: the mechanism is real, the front door is in progress, and neither claim is worth inflating.
The Better Version of This Story
Spirit's data found a price the week the company was already gone. The better version of this story is one where the people who create the value hold a share of it before anyone else can auction it. Sasha Mitchell, Elacity's founder, puts the reason plainly: 'The people who create the value should own it. That is the entire reason Elacity exists.'
Data is already capital. The only open question is whose, and that gets decided by whether ownership is built in before the sale or discovered at the auction. Follow Elacity on X for how the ownership layer gets built, one mechanism at a time.