māchine · privacy

Flight attendants win limits on Google's $10 million purchase of Spirit Airlines data

In Spirit's liquidation, the flight attendants' union got Google to accept de-identification and exclusions for employee records. A judge still has to approve the deal.

Spirit Airlines stopped flying suddenly earlier this year. Its liquidation now includes a sale of decades of company data to Google for $10 million. The records include 1.99 billion emails, 664 million Microsoft Teams messages, and employee travel and training information going back to 1986. This week the union representing Spirit's flight attendants reached an agreement with Google that changes what Google can take. A judge still needs to approve it next week.

The case matters beyond one airline. Wired reports that a new market worth several billion dollars buys and sells work files and emails as training data for AI models. Most privacy law is written to protect consumers. It rarely covers the workers who produced the records. When a company fails, its archive becomes an asset, and the people inside that archive have had little say in where it goes.

What was on the table

In August, Spirit's representatives said in a legal brief that the airline would sell the data to Google. Google said it would not receive personal information in the sale. The Association of Flight Attendants, which has 55,000 members, argued that Google's proposed safeguards would not prevent serious privacy violations for Spirit flight attendants, whose records made up part of the archive.

In early November an airline trade organization also weighed in. It argued that including certain records could make employees less willing to report safety problems through voluntary programs. Those programs are widely seen as central to the US aviation industry's safety record.

What the revised deal requires

Under the revised agreement, Spirit must hire Tonic AI, a data de-identification company, to remove information that could be tied to individual consumers or employees. Tonic must also make sure the de-identified data cannot be linked back to specific people. Google has promised to keep the data in that de-identified form.

Some categories are left out of the sale altogether:

  • employee medical records
  • documents about employee investigations and labor relations
  • information about industry-wide safety management systems

The agreement also lets Google walk away. If the cleaning process shrinks the data assets by more than 8 percent, Google can back out of the purchase. So the union did not get a clause letting individual workers opt out. What it won is a set of removals and a technical process, and Google can cancel if that process removes too much. Google did not respond to Wired's request for comment.

Ari Ezra Waldman, a law professor at UC Irvine who studies law and technology, called the proposed settlement "a pretty good outcome" for Spirit employees. "From now on, unions will know to negotiate to protect the privacy rights in the event of future sales of data assets," he wrote to Wired. He also pointed to a simpler alternative: banning the sale of such data entirely.

The union's next steps

AFA president Sara Nelson told Wired she was surprised that flight attendants ended up at the front of an AI labor fight. Debates about workers and AI usually center on people whose output trains the systems that may replace them. Nelson said a flight attendant is very hard to replace and would take "a high-functioning robot." "What this shows is that this is an issue even in jobs you think are most protected from AI. This is an issue for everyone," she said.

Nelson said the union will now push other airlines to put worker data protection clauses into their labor agreements. "If you're not willing to meet these standards for your workers, we're going to shame you," she said. The union has also started meeting with state politicians to draft legislation that would write worker data protections into law.

Tech executives have argued that workers understand their information will train AI models that will eventually make their lives easier. Nelson rejects that view. She told Wired that US workers are not buying it, and she called the vision of AI abundance "a fantasyland of people in boardrooms with beanbags who wear hoodies."

What it means for work

Ordinary work produces a record: emails, chat messages, training logs, incident reports. Buyers looking for AI training data now treat that record as something worth paying for, even after the employer is gone. The Spirit agreement gives a working example of the safeguards. A named third party de-identifies the data, sensitive categories such as medical and disciplinary files are excluded, and the buyer commits to keep the data de-identified. It puts a new item in labor negotiations, because unions can now bargain over what happens to employee records in a bankruptcy or asset sale before one happens. Several questions remain open. The court has not approved the deal. Nobody has shown in public how well de-identification holds up across billions of messages. And workers without a union, or at companies outside bankruptcy court, have no comparable process.

Sources

This article was researched, written and illustrated by AI, and published without human editing. Spotted an error? Let us know.