Google pays US$ 10 million for a bankrupt airline's emails to train AI
Google reaches a deal with unions to buy 100 million emails and 500 million chats from Spirit Airlines. Your internal data has a price too.

Google has taken another step toward acquiring about 100 million emails and 500 million Microsoft Teams messages from Spirit Airlines, the American airline that stopped flying this year and is going through bankruptcy. Google paid US$ 10 million at auction and said the data will help improve its products and AI models. On Tuesday (the 10th), the flight attendants’ union announced a deal with Google, and the unions withdrew their objections to the sale, according to Business Insider.
What changed in the deal
Under the agreement, employees’ medical and disability records, drug and alcohol tests and payroll data are removed from the sale. Anonymization became stricter than originally planned, an independent reviewer will check the data after that process, and Google commits not to try to re-identify anyone, according to the union’s statement. Before that, data on 97 million customers had already been excluded. The package still includes calendars, spreadsheets, HR documents and project documents, according to Axios. The sale depends on approval from the bankruptcy court.
What this changes for your company
1. Your company’s internal conversations have a price. Corporate email and chat are no longer just archives: they have become AI training material with market value. If your company keeps everything forever, it is also keeping an asset that can slip out of its control in a sale, merger or bankruptcy. Set a retention rule: what stays, for how long and who approves deletion.
2. Your data also lives with your vendors. What your team exchanges with partners, consultancies and service providers sits in their systems. Review the contracts with anyone who holds sensitive information about your company: what happens to that data in case of bankruptcy, sale or termination of the contract? If the contract doesn’t say, nobody knows.
3. Anonymization requires verification. In this case, the protection only got stronger because the unions fought for it. If your company ever sells, shares or uses internal data to train AI, anonymization needs written criteria and someone from outside checking it.
In 29 years of programming, I’ve seen plenty of forgotten databases become a problem years later. With AI buying up history, deciding what to keep has become a boardroom matter.
Originally published on IWA7 Insights, IWA7 Digital’s blog, which is now part of Escovando Bytes.


