Key facts
- Google won an auction to acquire a dataset from Spirit Airlines for $10 million.
- The data includes employment and workplace records, not personal or customer information.
- Google agreed to de-identify the data before transfer, with a third party overseeing the process.
- The Association of Flight Attendants (AFA) objected, citing a lack of protection for employee data.
- A US court has delayed a hearing on the data purchase due to the union's objections.
A US court has delayed a hearing concerning Google's $10 million acquisition of data from the bankrupt Spirit Airlines, following objections from the Association of Flight Attendants (AFA). The union, representing Spirit workers, argues that the deal, which includes extensive employment and workplace records, offers insufficient protection for former employees.
Google won the auction for the dataset, which spans decades and includes approximately 100 million employee emails, HR information, and payroll data. The tech giant agreed to have a third party scrub the data of personally identifying information (PII) and committed to not intentionally re-identify individuals. However, the AFA contends that consumer protection laws, which Google cited, do not adequately cover worker confidentiality, creating a loophole.
The union fears that even de-identified data could be combined with other datasets by Google or third parties to re-identify individuals. They argue that sensitive employment information, such as disciplinary records or scheduling grievances, remains confidential regardless of whether an employee's name is stripped from it. The Electronic Frontier Foundation also expressed concern, stating that using employee data for AI training without consent is problematic.
Spirit debtors prioritized buyers with plans least likely to frustrate customers, and Google's bid was favored for its price and commitment to data scrubbing. Mercor Corporation was a competing bidder, offering a similar deal that would be accepted if Google fails to complete the purchase.
