AI companies and data brokers have sought copies of workplace documents, emails, Slack messages, support tickets and call transcripts for AI training and testing, according to WIRED. In its November 6, 2026, report, WIRED said a proposed sale of Spirit Airlines’ digital archive was on hold after the flight attendants’ union objected to workers’ private information being included.
The dispute points to two separate routes by which employee data can reach AI systems: an employer can use its own workforce records internally, or a company can license or sell records to an outside buyer. Both raise questions about what the records contain and how identifying details are handled.
Removing names does not necessarily remove the personal details embedded in a work conversation.
What records are being sought
The pitches described by WIRED covered the everyday paperwork of an organization: documents, emails, Slack messages and tickets. One pitch also named call transcripts. Handshake told WIRED that it licenses professional documents for AI testing and training; WIRED also described Mercor and Micro1 as data-acquisition firms.
The sums in these accounts refer to distinct reported offers, not a standard rate. An unsolicited pitch to WIRED offered a one-time $350,000 to $1.8 million for a copy of old company records. WIRED separately reported that Handshake offered a freelance writer $1,000 for notes from one story; she declined. Business Insider reported that Handshake AI offered up to $30,000 for high-quality written documents, on the condition that the submitter owned them and had authority to share them. That reported ceiling does not establish what contributors typically receive.
A brokered sale is different from internal use
A data broker or licensing company arranges for records to reach an outside organization. Business Insider reported a separate path at SpaceX: Elon Musk told employees their data would be used to train Grok. That was described as employer use of employee data, not a third-party purchase.
The scope of any company’s practice matters. Anthropic told WIRED it does not buy workflow data from live companies such as WIRED. That statement concerned this kind of company-workflow data, rather than every possible category of data.
Spirit Airlines’ archive put worker privacy at issue
WIRED reported that Spirit Airlines’ archive included 80,000 email accounts, 500 million internal messages, 20.6 million SharePoint files and 1 million time-card records. The flight attendants’ union objected to workers’ private information being included in the proposed sale. WIRED reported on November 6, 2026, that the sale was on hold.
The scale helps explain why a company archive is different from a single document submission: email and workplace messages can contain information about many employees, across many parts of their working lives. The counts describe the records WIRED reported in Spirit’s archive; they do not, by themselves, say what personal details each file contains.
What removing identifiers can leave behind
Mercor CEO Brendan Foody told WIRED that the company removes personally identifying information from business data it receives. EPIC general counsel Mayu Tobin-Miyaji raised a separate concern: unstructured emails and Teams messages can include sensitive details, such as health appointments or family matters, that may be harder to identify and remove.
Removing a name or other direct identifier is therefore not the same as showing that every sensitive detail has been removed. The concern is about what may remain inside the text of workplace communications, not evidence that a particular record in Spirit’s archive exposed such details.
