Google wins $10M auction for Spirit Airlines' business data to train AI

The data haul includes about 100 million company emails, 500 million Microsoft Teams chats, 30 million lines of code, software algorithms, development metadata, and records related to the airline's revenue, aircraft operations and employee productivity, Bloomberg reported. Google outbid several competitors, including the AI training data company Mercor.io Corp.

The information does not include Spirit's 97.5 million passenger profiles or its 50 million loyalty program records. Google said it will ensure that all acquired data is "rigorously scrubbed" of personally identifiable information by a third-party service provider before Google accesses it.

Google said the data is a valuable asset that will help improve its AI products and models. The purchase reflects a broader trend in which proprietary business data has become a sought-after resource for AI companies seeking to train more powerful models. A nascent market has emerged for failed or bankrupt companies to sell their emails, chat logs and operational data to AI firms, as Gizmodo reported in April. Mercor, which lost to Google in this auction, has been offering professionals cash for materials from former employers to extract industry-specific knowledge for AI training, according to a Wall Street Journal report that month.

AI companies are willing to pay large sums for such data because they need far more information to develop models that can outperform humans. Google, OpenAI and Anthropic have already consumed much of the freely available data on the internet, and experts have warned that the world is running low on fresh data, which could hinder future improvements to AI systems.