AI Firms Battle for Spirit Airlines Internal Data

Google secured an initial bid of USD 10 million for internal data from Spirit Airlines. This archive includes 100 million emails, 500 million Microsoft Teams messages, 30 million lines of code, and employee records dating back to 1986. AI startup Micro1 subsequently offered USD 12.5 million for the same collection, prompting a judicial review.

Credit: UNSPLASH

Micro1, an AI training organisation led by Chief Executive Officer Ali Ansari, submitted its revised bid after the initial deadline. The firm argued that decades of authentic operational records are essential for developing advanced AI systems. Ansari contended Google's original valuation was insufficient for such material. A ruling on the matter is expected on Sept. 9.

And this situation originated from the bankruptcy auction of Spirit Airlines' archives. The airline ceased operations earlier this year after its second bankruptcy, rendering over 17,000 workers redundant and accumulating debts of approximately USD 8.1 billion. Its extensive internal communications, previously unvalued, became a contested asset. Google was the initial winning bidder earlier in Aug.

Micro1's subsequent offer has made the final status of Google's bid uncertain. The sale excludes passenger profiles and frequent flyer accounts but includes records of former employees. Companies now seek such private archives as publicly available internet data for AI model training becomes scarce due to limitations in both volume and scope.

But according to BTUAI, only about 15 per cent of global knowledge is digitised and searchable. Public internet information typically provides outward facing statements, rarely revealing granular internal operations. An archive such as Spirit's captures work as it unfolded, offering real world data like revenue team debates and maintenance issue escalations.

This content, documenting decisions and mistakes, is considered more valuable for training enterprise level AI. According to Bloomberg Law, Google intends to use this data to improve its products and AI models. A privacy measure is incorporated into the data transfer process.

So, a third party will undertake deidentification of the archive before Google gains access, stripping out personal details that directly identify individuals. Google has formally agreed not to reverse this deidentification process. However, a specific condition mandates that links between records must remain intact.

This preservation allows an employee's activity to be tracked across emails, chats, and files, a feature precisely valued for AI training. These maintained connections concern the flight attendants' union, which formally objected to the sale, arguing individuals within the 17,000 strong workforce could still be reidentified.

And the bankruptcy judge postponed the approval hearing to Sept. 9. Consequently, Google has won the bid but not secured the data. This year has seen a market emerge for the data of defunct companies, with startups selling internal records to AI firms for creditor recovery. Bankruptcy law, codified in 1978, treats data as a business asset.

But this legal framework predates modern understanding of data's value and privacy concerns. The upcoming hearing will address the union's privacy objection and Micro1's additional USD 2.5 million offer. Spirit Airlines' demise highlights a need for organisations to review data retention policies, privacy safeguards, and employee awareness regarding communication persistence.

So, employment agreements and vendor contracts defining data ownership require review. Initial AI data disputes focused on public web content; future conflicts are anticipated over private corporate archives. This case raises pertinent questions for all organisations regarding data protection in the evolving AI data environment.

• Google initially bid USD 10 million for Spirit Airlines' internal data.

• AI startup Micro1 made a late, higher offer of USD 12.5 million for the same data.

• The data includes emails, messages, code, and employee records from the defunct airline.

• A privacy protocol involves third party deidentification, but links between records will remain intact.

• The flight attendants' union has objected to the sale due to reidentification concerns.

Source: Forbes