Google bid $10m for a dead airline's inbox. That is now the comp for your internal records

Price per item is the figure worth keeping. Add the emails, the Teams messages and the lines of code and the archive runs to about 630 million discrete records, which puts the winning bid at roughly 1.6 cents each ^[2]. Divide the same $10 million by the 17,000 people who lost their jobs when Spirit stopped flying on May 2 and it comes to about $588 a head ^[3]; Micro1's late offer would raise that to about $735 ^[3]. Nobody in that group negotiated the rate. They produced the corpus as a byproduct of doing their jobs, across roughly four decades of records ^[4].

Passenger profiles and frequent flyer accounts were carved out of the sale. Employee records were not ^[6]. That asymmetry is the lesson: consumer data has had statutes and lawyers behind it long enough for an estate to know it was dangerous to sell, and workplace exhaust has not.

The privacy remedy on the table shows why the distinction bites. A third party de-identifies the archive before Google receives it, stripping names and addresses, and Google agrees not to reverse the process ^[7]. But the sale agreement requires the links between records to stay intact ^[8], because following one worker's thread across email, chat and files is the thing being bought. The flight attendants' union objects that patterns alone could re-identify people inside a 17,000-person workforce ^[9]. Both things can hold at once, and that is the difficulty: the buyer's specification and the privacy fix ask the same dataset for opposite properties.

Underneath sits a mechanical fact with nothing to do with AI. Bankruptcy treats data as an asset of the business, ranked next to a gate slot or an aircraft ^[10], under a code written in 1978 ^[11]. That code also leaves room for a late bid when it puts more money in creditors' pockets, though judges rarely re-open a closed auction ^[12]. The September hearing therefore turns on recovery arithmetic as much as on privacy.

Forbes reports that a small market formed this year around the records of failed companies, with startups selling internal messages to AI firms so creditors can recover something ^[13]. Google, per Bloomberg Law, says the data will improve its products and AI models ^[15]. The sell-side framing is scarcity: BTUAI's estimate that only about 15 percent of the world's knowledge has ever been digitised ^[14] is that argument in one number. The operator framing is narrower and less comfortable. A retention policy that keeps Teams logs indefinitely is accumulating an asset whose eventual buyer gets chosen by a creditors' committee, not by the executive who wrote the policy.