Google has agreed to purchase a decade-long data archive from Spirit Airlines for $10 million as part of the carrier's bankruptcy proceedings. The acquisition scope includes 1.99 billion emails, 664 million Microsoft Teams messages, and detailed logs of employee training and travel dating back to 1986. The budget carrier filed court documents regarding the pending deal in August, accompanied by assurances that personal data would be excluded from the transaction. The $10 million price tag for nearly two billion emails and hundreds of millions of messages establishes a concrete secondary market benchmark for corporate communications during asset liquidation: roughly half a cent per unit of unstructured content prior to sanitization. However, the Association of Flight Attendants union, representing 55,000 members, stated that the protective measures proposed by Google are inadequate. The union's stance turns the acquisition of corporate databases into a contentious dispute ahead of next week's court approval hearing. The central question for the buyer remains pragmatic: how much useful signal will remain in these datasets after de-identification?
Audit trails and the cancellation threshold
Spirit Airlines is required to retain Tonic AI, a firm specializing in data de-identification. Tonic AI engineers must excise information that could potentially be linked to specific individuals and guarantee that the sanitized data cannot be re-identified. For its part, Google committed to storing the acquired data exclusively in this anonymized format. Furthermore, numerous sensitive documents will be excluded from the sale, including employee medical records, internal investigations, labor disputes, and information regarding industry safety management systems.
"Now unions will know that when asset sales include data, they need to push for privacy rights protections," noted Ari Ezra Waldman, a law professor at the University of California, Irvine.
However, technical sanitization introduces financial risk to the transaction. Under the agreement terms, Google can cancel the purchase if the data sanitization process reduces the asset volume by more than 8%. Any excessive redaction of logs or removal of disputed fragments can instantly render the asset legally worthless and deprive the carrier's creditors of the promised $10 million. This exposes the underlying vulnerability of such investments: models require text for training, but stringent regulatory and union demands to purge personal data can destroy the dataset's utility before the deal even closes.
Conflicting arguments and industry barriers
Tech executives maintain that employees understand their data will be used to train artificial intelligence models, which will ultimately streamline their work. The union side rejects this argument outright.
"Replacing a flight attendant is exceedingly difficult—it would require a highly functional robot, and we are a long way from that," says Sara Nelson, president of the Association of Flight Attendants. She notes that the situation demonstrates how the issue affects even those professions previously considered insulated from AI disruption, impacting everyone.
The flight attendants' position has found support among airline industry associations, which warned that including such data in asset sales could discourage employees from reporting safety issues through voluntary reporting programs—mechanisms considered critical to maintaining high safety standards across the U.S. aviation sector. Professor Waldman points to another, potentially simpler alternative: an outright ban on the sale of such data.
The Spirit Airlines case demonstrates that the era of acquiring corporate logs through bankruptcy proceedings is hitting structural resistance. If Tonic AI's external filtration strips away more than 8% of the acquired volume, the technology giant will walk away from the contract, proving that the real market value of toxic internal datasets stripped of personal data trends toward zero.
What doesn’t add up
Bankrupt Spirit Airlines secures an eight-figure payout for its communications archive, yet the buyer will walk away if data scrubbing wipes out just eight percent of the material. That strict threshold forced on Tonic AI turns millions of internal logs into an all-or-nothing technical gamble. Meanwhile, pushback from flight attendants proves that corporate archives are valuable precisely until realistic working context gets stripped away. Buying bankrupt corporate databases has hit an impasse: without personal context, such training data loses its entire commercial utility.