Google buys Spirit Airlines employee data for AI training

Google secured a massive dataset from bankrupt Spirit Airlines with a winning bid of $10 million during a virtual auction on August 14. The tech giant outbid Mercor Corporation, which placed a backup bid of $7.5 million. Google started the bidding at $5 million and agreed to cover the costs of a third-party service to scrub personally identifying information from the records. The acquired assets include Spirit's computer programs, applications, code, and decades of worker data, including approximately 100 million employee emails, payroll records, and productivity metrics. Google plans to use this information to train its artificial intelligence models.

However, the Association of Flight Attendants filed a limited objection to the sale. The union argues that while consumer-facing data like customer profiles and loyalty program details were excluded, sensitive employee records remain highly vulnerable. Although Google has committed to a court-appointed ombudsman and promised never to intentionally re-identify the scrubbed data, the union warns that de-identification only strips names. It does not protect the confidentiality of internal communications, disciplinary records, or training deficiencies, which could still be linked back to specific groups or individuals.

For AI practitioners and developers, this acquisition highlights the evolving legal and ethical landscape of sourcing training data from bankrupt enterprises. While using corporate archives provides a rich, real-world dataset for training enterprise AI, it introduces significant compliance risks. Practitioners must navigate the distinction between consumer privacy laws and worker confidentiality. As unions push for stricter guardrails, AI developers may face tighter restrictions on how they process, segment, and utilize employee-facing data, even when it has been legally acquired and technically anonymized.