Memorandum
- From
- Derek Weston via Fast Company
- Date
- Filed
- Business·4 min to read
- Re
Spirit Airlines’ grounded data draws rival AI bids after Google’s $10 million offer
ReSpirit Airlines’ grounded data draws rival AI bids after Google’s $10 million offer
Spirit Airlines, grounded since May, is selling its corporate data for AI training. Google bid $10 million, but rival Micro1 has offered $12.5 million, sparking a bidding war over a new asset class.
Spirit Airlines, which halted operations in May amid bankruptcy proceedings, is now at the center of a bidding war over a surprising asset: its internal corporate data. Google has successfully bid $10 million for the dataset, which includes internal wikis, emails, source code, and spreadsheets, according to court documents. However, AI training provider Micro1 has since submitted a rival offer of $12.5 million, though it remains unclear whether the late bid will be considered.
The sale of Spirit’s data highlights a growing trend where companies building and refining AI systems acquire operational records from both shuttered and active businesses. These datasets, which can include Slack chats, GitHub logs, and customer support documentation, are used to train AI models to handle office work in realistic business environments. A Google spokesperson confirmed the acquisition, stating, “We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models. We will not receive any personal information from this dataset.”
Under the Google deal, any personally identifiable information would be processed by a third-party “deidentification agent,” and the dataset does not include Spirit’s customer lists. However, a Spirit flight attendant union has formally objected to the sale unless more steps are taken to protect employee privacy, arguing that workers can still be identified even after their names are removed. Privacy advocates echo these concerns, warning that sensitive information could resurface in AI outputs despite anonymization efforts.
The bidding war for Spirit’s data is part of a broader rush by AI companies to acquire diverse datasets, from Reddit’s forum archives to out-of-print books. Mercor, another AI training company that offered $7.5 million for the Spirit data, noted, “Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI.”
Micro1, which operates a program called Data Partnerships, uses such corporate records to build simulated work environments for reinforcement learning. Founder and CEO Ali Ansari explained that AI agents are set up as employees at fictitious companies generated from anonymized data, equipped with realistic software tools to complete tasks. “It needs to have a world that it refers to, whether it’s company files, Slacks, documents, etc., so that it can do realistic actions and refer to realistic data,” Ansari said.
Before the data is used, Micro1 strips out sensitive information through an AI-driven process, and raw data is typically deleted within 30 days. Identifiers like email addresses are replaced with simulated dummy data. The company also tells data providers not to share trade secrets, legally privileged files, or health data, and excludes such information from training sets when it appears.
The sale of corporate data from bankrupt companies is not new, but the growth of AI has created new demand for operational records that previously had little market value. Alexandra Mateescu, a researcher at Data and Society, noted, “Particularly now that these big AI companies are just increasingly hungry for workplace data to train their AI models, there’s just a massive expansion of what counts as commodifiable data.”
