Google Buys All of Spirit Airlines' Data to Feed Its AI Models
- Reference: 0185039636
- News link: https://tech.slashdot.org/story/26/08/18/1652202/google-buys-all-of-spirit-airlines-data-to-feed-its-ai-models
- Source link:
> Did you ever fly on Spirit Airlines? Or work there? Or send an email to someone who worked there? Then your information will soon be feeding Google's artificial intelligence model. The discount airline halted all operations in May, and it's been selling off its remaining assets through the bankruptcy process ever since. While most of that is planes, equipment, and real estate, there's other things that are valuable, too.
>
> It was disclosed late Monday that the company agreed to sell its data to Google for $10 million. That includes emails and internal communications, spreadsheets, transactions with the public including bookings and frequent flyer information as well as human resources information on its employees. The data has been stripped of anything that would allow individuals to be identified, according to the court filing. But it is still a massive amount of useful intelligence.
[1] https://www.cnn.com/2026/08/18/business/google-spirit-airlines-data
Can't wait for a story about failed anonymization (Score:3)
You know it's going to happen - someone in charge of replacing/stripping all the names and replacing them with numbers, or whatever, is going to fail at some level. A bunch of PII will go into Gemini, and they won't be able to sue Spirit, because they won't exist, and Google will definitely make sure their contract excuses them from any liability.
Re:Can't wait for a story about failed anonymizati (Score:4, Insightful)
> You know it's going to happen - someone in charge of replacing/stripping all the names and replacing them with numbers, or whatever, is going to fail at some level. A bunch of PII will go into Gemini, and they won't be able to sue Spirit, because they won't exist, and Google will definitely make sure their contract excuses them from any liability.
Even if it would be truly anonymized, once combined with th other vast amounts od data Google has means it is possible to tie the anonymized data to specific, or a plausible group of people.A bad anonymizing process would be a far more likely culprit, but by analyzing patterns, something AI and computers are good at, could yield interesting results.
Re: (Score:3)
> Even if it would be truly anonymized, once combined with th other vast amounts od data Google has means it is possible to tie the anonymized data to specific, or a plausible group of people.A bad anonymizing process would be a far more likely culprit, but by analyzing patterns, something AI and computers are good at, could yield interesting results.
It has already been studied and accomplished. Too many studies to name here, but for example, in 1997 a Massachusetts researcher succeeded at converting zip code, birth date, and gender info from voter-registration lists, into links to someone's medical records.
Re: (Score:2)
> You know it's going to happen - someone in charge of replacing/stripping all the names and replacing them with numbers, or whatever, is going to fail at some level. A bunch of PII will go into Gemini, and they won't be able to sue Spirit, because they won't exist, and Google will definitely make sure their contract excuses them from any liability.
It doesn't even have to fail. I would bet Google has enough other data that they use re-associate at least some of this data. How many people use Google services with their flight (or hotel) information for alerts, calendar, flight tracking, etc...
Re: (Score:2)
Most of the risk of de-anonymization comes from "enrichment" - you may not have a name in this dataset, but say you have birthdate, gender and zipcode. In that dataset you have names, and you match them up based on the other attributes.
If you read claims about anonymized data closely, you'll notice the providers are well aware of this in how they phrase things.
Of course plain old incompetence also does the trick.
Re: (Score:2)
For example, on August 18th there's an email/message from AnonymisedSpiritEmployee_0175034 saying, "Thanks for the cupcakes today and to everyone who contributed to my birthday gift card to Benihana! My husband and I eat there every month for date night. Yummy!"
From one message you have birthdate, probable gender, marital status, and spending activity which can be trivially cross-referenced against existing data collation from all kinds of other sources. Sources that Google also owns.
Now We Know (Score:3)
We now have a NUMBER for the value of these data breaches. Time for the Lawyers to use this metric to go after every single data breach.
Re: Now We Know (Score:2)
I guess it's also a price on the "value" of your individual data. Which is about $1, maybe. At most.
Re: Now We Know (Score:4, Informative)
Google AI tells me "Prior to shutting down on May 2, 2026, Spirit Airlines accumulated roughly 97.5 million customer profiles in its database and about 50.2 million members in its Free Spirit loyalty program". So my $1 was an overestimate by about 1 order of magnitude. 10c is all you're worth!
Re: (Score:3)
> Time for the Lawyers to use this metric to go after every single data breach.
Hardly. Unfortunately not all data breaches are the same, and neither is the value for people. The value of data to someone who is using it for extortion is different than for someone using it for impersonation, is different from someone using it to feed it to their AI.
$10million isn't an open market price. It's what one company thought *that* data was worth to them.
Re: (Score:2)
You got my hopes up thinking you were talking about a Mitchell and Webb sketch.
Okay, but (Score:5, Funny)
Knowing Spirit airlines, the data will be delivered six hours late, to the wrong destination, missing its codebook (which somehow got rerouted to Dallas), and the data will be hot, tired, thirsty, and extremely pissed off.
I pity the fool who does business with Spirit.
Re: Okay, but (Score:2)
Spirit is kaput. Google is picking up wreckage from the scene at scrap value.
But you do make a point that most people underappreciate. The Spirit airline experience was bad, but one of the bigger factors in the poor experience was that everyone on the planes were extremely grumpy. You could avoid being nickel and dimed but there was no escape from the foul mood. No other airlines have ever come close.
Teaching AI to fail (Score:1)
If spirit airlines failed, why waste resources learning their crap at all? We already know spirit's technique is the wrong way to do it.
Re: (Score:2)
> If spirit airlines failed, why waste resources learning their crap at all? We already know spirit's technique is the wrong way to do it.
*sigh*
Because Spirit Airlines failed at BEING AN AIRLINE. Google isn't trying to use the data to become an airline, they're adding it to the meat grinder with which to train Gemini.
Re: (Score:2)
They're not teaching an LLM how to manage airline commercial strategies. They're teaching it to write text, read text, understand spreadsheets, etc. You can do that fine on airline's internal communications, even an airline that went bust.
Re: (Score:3)
> They're not teaching an LLM how to manage airline commercial strategies. They're teaching it to write text, read text, understand spreadsheets, etc. You can do that fine on airline's internal communications, even an airline that went bust.
You don't think Google's B2B divisions are interested in selling AI-powered BookingasaService, GateCapacityManagementasaService, MaintenanceSchedulingasaService, FlightCrewPayMinimizationasaService, etc. to other transportation companies?
Hoillywood ... (Score:2)
> If spirit airlines failed, why waste resources learning their crap at all? We already know spirit's technique is the wrong way to do it.
Hoillywood, AI generated airports scenes. They just need to look realistic, meybe even better if a little flawed and comical.
Anonymized yet Correlatable? (Score:2)
Even if obvious PII is removed from this data, that's not likely to add much privacy for many people - esp. employees and regular customers.
This is in part because things like patterns of email would, with fairly high probability, identify some employees in key roles - even without attempting to correlate it with external data sources.
For example using just email where pseudonyms for each person and email address replaced actual names and email addresses (both in metadata and email bodies), one could probab
Re: (Score:2)
Not saying you're wrong..... but
Why would you generate pueudonyms? Just delete it. Delete the To: From: fields for all email.
Delete the passenger name/number from every listing. This isn't information that's going to be used to follow trends, debug software, or optimize an airline. Instead of 'anonymizing' by replacing or obscuring. Anonymize it by simply blanking it out.
There's no need for them to know that 'Person A' communicated with 'Person B' this many times. It's not that kind of data ana
Re: (Score:2)
There's really no "need" for Google to have any of this information (they wouldn't have had it if Spirit had been purchased and Google wouldn't have gone out of business as a result).
The article didn't say what Google wants to use the data for.
The data would likely have some value even with removal of all names/email addresses/employee numbers/etc -- i.e., absolutely no way to correlate any record with any other record. Even dates of hire and termination in the HR records could help identify the associated
The new AI (Score:2)
In other news, Google is working on an AI model that can de-anonymize anonymized data sets.
slashdot wants to make millions? (Score:2)
Does slashdot want to make millions on selling its data by any chance? And let me just say I'd love to get at least 0.01% off the deal for the idea.
No such thing as anonymous datasets. (Score:3)
> The data has been stripped of anything that would allow individuals to be identified
We all know this is not possible. This information will be identifiable by correlating with other information, including public information.
Privacy Agreement, What Privacy Agreement? (Score:2)
Anyone see Google's intended use of the collected data spelled out or even implied in the "Primary Purpose for Collection and Use of Data" or elsewhere in the [1]Spirit Privacy Agreement [spirit.com]? Asking for 1.2 billion Spirit passengers.
[1] https://content.spirit.com/Shared/en-us/Documents/Privacy_Policy.pdf
So let me get this straight... (Score:3)
So.... Google is using data from a bankrupt airline that had customer service and reliability problems through most of its life?
And people are going to be using AI trained on that data?
What's their motto, "Share and Enjoy"?
Re: (Score:2)
It just means Gemini will start being rude and will refuse to try to help anyone with a question.
Re: (Score:2)
At a minimum, I'm guessing.
Re: Anonymized (Score:2)
They _may_ already have a lot of it anonymised for use with internal tooling. Realistically, though...
Re: Anonymized (Score:2)
Thank you for your feedback *aniels. We'll get back to you at 2?2-555-1212 in the next couple of days.
Will be useful for AI generated comedies (Score:2)
Even anonymized data should be pretty entertaining. Spirit airlines trained AIs depicting brawls in the background of airport scenes, lots of dildos at luggage screening depictions. These AI will be in high demand for comedies.. :-)