Google and the flight attendants' union agree on de-identification of data from the Spirit Airlines sale
Google will buy ten years' worth of work records from the sale of Spirit Airlines' assets for 10 million dollars. Following objections from the flight attendants' union, Spirit Airlines must first de-identify the data via the service Tonic AI, sensitive records are excluded, and Google can withdraw from the purchase if the data volume drops by…
As part of the liquidation of Spirit Airlines, which suddenly ceased operations this year, Google agreed to purchase ten years' worth of work data for 10 million dollars. This involves 1.99 billion emails, 664 million Microsoft Teams messages, and employee records dating back to 1986. According to Spirit Airlines, Google was supposed to receive the data without personal information, but the flight attendants' union, the Association for Flight Attendants (55 000 members), objected that the originally proposed safeguards would not prevent serious violations of employees' privacy.
Following negotiations, a revised agreement emerged: Spirit Airlines must hire a third party, the service Tonic AI, to de-identify the data so that it cannot be traced back to specific individuals, and Google has committed to retaining it only in this form. Sensitive categories of records are excluded from the sale — employees' medical records, materials from internal investigations and labor disputes, and information on safety reports within industry safety reporting systems. The aviation industry association noted in November that including the latter data could discourage employees from voluntarily reporting safety issues.
The agreement also includes an opt-out clause: Google may withdraw from the purchase if the de-identification process reduces the data volume by more than 8 percent. The agreement still needs to be approved by a court next week. The union's president, Sara Nelson, stated that the organization will now seek similar protective clauses on work data in collective bargaining agreements with other airlines and is in talks with state-level policymakers about legislation to protect employee data. Legal scholar Ari Ezra Waldman of UC Irvine assesses the agreement as a good precedent for future union negotiations, while also noting that an outright ban on selling such data could be an alternative.
Why it matters
The case shows that employees' work emails and internal messages are routinely sold as a source of data for training AI models, even in professions previously seen as "safe" from automation. For employees, this means their work communications could end up in datasets without their direct consent unless a collective bargaining agreement explicitly addresses this. For companies buying and selling data, a precedent is emerging whereby unions can negotiate mandatory de-identification, exclusions of sensitive categories, and opt-out thresholds based on data volume, which increases the legal and procedural requirements for such deals.
Two audiences, two different impacts
What this means
For individuals
Employees at unionized workplaces can demand that collective bargaining agreements include protection of work data against sale to third parties for AI purposes, similar to what the Association for Flight Attendants is now pushing for.
For a business
Companies selling or buying work data for AI training must take into account that labor unions can negotiate mandatory third-party de-identification, exclusions of sensitive record categories, and contractual thresholds (e.g., an opt-out if the data volume drops by 8%), which affects the cost, scope, and legal risk of such deals.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.