Business & Finance
Flight attendants union opposes Google's $10 million bid for bankrupt Spirit Airlines data
The Association of Flight Attendants argues the proposed sale of 34 years of company data to train AI models threatens former employees' sensitive personal information.
The short version
- Google won a $10 million bankruptcy court bid to purchase 34 years of Spirit Airlines operational and internal data for AI model development.
- The Association of Flight Attendants filed a legal objection, contending the dataset includes sensitive employee medical, financial, and personal records that lack adequate consumer-level privacy protections.
- Google stated that the dataset excludes customer data and that employee records will be deidentified by a third party.
- A bankruptcy court hearing regarding the proposed sale is scheduled for September 9.
Key facts
- Google submitted a winning $10 million bid in bankruptcy court to acquire 34 years of Spirit Airlines data for AI training, topping a $7.5 million bid from Mercor.[Wired]
- The Association of Flight Attendants, representing 5,500 former Spirit employees, filed a formal objection to block or restrict the sale.[Wired]
- According to court documents, the proposed sale encompasses over 175,000 employee records, 1 million time cards, 80,000 email accounts, and hundreds of millions of Microsoft Teams, OneDrive, and SharePoint files.[Wired]
- Google stated that no consumer data is included in the purchase and that a third party will strip identifying elements from the data.[Wired]
- The union is separately seeking approximately $68 million in unpaid wages, healthcare costs, and vacation pay from the airline's bankruptcy estate.[Wired]
What remains uncertain
- Whether the bankruptcy judge will approve the transaction, order specific employee data to be excluded, or mandate enhanced privacy protections at the September 9 hearing.[Wired]
- Whether standard de-identification procedures will successfully prevent personal information within internal files and emails from being re-identified by AI models.[Wired]