Data brokers pitch Wired to sell old work emails to AI labs

A Wired author describes a new kind of cold pitch landing in the inbox. Someone named Nick wrote that AI labs such as Anthropic and OpenAI are paying media companies such as Wired $350K to $1.8M as a one-time fee for a copy of their old work records: documents, emails, Slack messages, tickets. A second email, which added call transcripts to the wish list, raised the possible payout to $2 million. A third writer, Cora, claimed that public data is exhausted and that labs are spending over $45B on private data. These figures come from the pitch emails alone.

Looking into it, the author found a business much bigger than expected, one that its practitioners say could affect workers and society as much as the data centers dominating the news. A freelance writer wrote about Handshake offering her $1,000 for her work on a single story; she did not sell. A Handshake representative confirmed the offer: the company licenses professional documents across many fields so labs can test and train how models handle real knowledge work, and journalism is one of those fields. Handshake also says a big part of its business is buying graduate students' papers and coursework. Another company, Sunset, writes to startups offering to help them wind down, scrub and anonymize their data, and broker the deal with the labs.

The industry leaders, Handshake, Micro1 and Mercor, are only a few years old and are snapping up databases like a modern gold rush. Mercor, founded in 2023, seems to lead; last year its three cofounders became billionaires at 22. Like the others, it claims partnerships with just about all the major AI labs. The labs are more circumspect: Anthropic acknowledges a big market for data but says it does not buy workflow from live companies like Wired. OpenAI gave no response before publication.

Mercor CEO Brendan Foody told the author the market is poised to explode. Mercor began as an AI recruiting company, then learned that big labs were frustrated with data quality: hired experts simulating workflows did not capture the messiness of a real operating business. He said Big AI has two bottomless needs, compute and data. On privacy, he said Mercor strips personally identifying information from data a business streams to it, and that this de-identification can be verified. Micro1 CEO Ali Ansari, who founded the company in 2022, said labs, enterprises and robotics companies pay about $40 to $50 billion a year for data (the author could not verify this) and predicted over a trillion dollars a year in two to four years. Micro1 mostly uses what it buys to improve its own models but also sells to big AI firms. Ansari said about half of YC startups are data companies; the author corrects this: in the most recent batch, 40 of 65 launched companies are classified as data startups.

The bankrupt Spirit Airlines put its digital assets up for sale: 80,000 email accounts, 500 million internal messages, 20.6 million Microsoft Sharepoint files and a million time-card records. Google seemingly won with a $10 million bid, then Micro1 topped it by $2.5 million, setting off an argument over who the winner was. The sale is now on hold because the flight attendants' union has charged that it involved their private information. The Electronic Privacy Information Center has filed an amicus brief for the flight attendants. Its general counsel, Mayu Tobin-Miyaji, said it is not clear anonymizing can fully remove privacy issues, since emails and Teams messages hold unstructured data, such as talk of doctor's appointments or family issues. Venture capitalist Gordon Ritter, who has taken up the issue as a cause, said norms and standards for data acquisition have not been set and that the AI industry is moving far faster than the ability to set them, partly on purpose.

The author calls the most troubling aspect that the movement enlists workers, not always with explicit consent, in an effort whose ultimate goal is to replace or automate the human work behind the documents. The flight attendants' union president told an interviewer the data only became valuable in the age of AI in order to train models to replace jobs. Asked whether selling workflow data will become the corporate default, Ansari said yes, and that it will become more common than ads as a revenue line item. He conceded that pushback against AI is a fair point but thinks objections will be manageable, because workers will see the upside of better models.

The author also followed up on one cold pitch, from a woman with a gorexen.com address, and met Jonah Stillman, head of commercialization at Tenkara, a two-year-old company. Tenkara wants data to improve an AI model that runs office operations for manufacturing companies; how Wired's archives would help is, the author says, a mystery. The author is not authorized to sell the documentation of Wired's editorial process and did not offer it. The piece is an edition of the Backchannel newsletter by Steven Levy and Lauren Goode.

Key facts

• Cold emails claim AI labs pay media companies $350K to $1.8M, and possibly $2 million, for old work records such as emails, Slack messages and tickets; the figures come from the pitch emails alone.
• Handshake, Micro1 and Mercor lead a young industry buying workplace data. Micro1 CEO Ali Ansari says labs, enterprises and robotics firms pay about $40 to $50 billion a year (unverified) and predicts over a trillion dollars a year in two to four years.
• Anthropic says it does not buy workflow from live companies like Wired; OpenAI did not respond before publication.
• The sale of bankrupt Spirit Airlines' digital assets (80,000 email accounts, 500 million internal messages) is on hold after the flight attendants' union charged it involved their private information.
• Critics, including EPIC's general counsel, doubt anonymizing removes all privacy risk, and the author sees the larger issue as workers feeding systems meant to automate their jobs.

Why it matters

The piece argues that data, not only compute, is a bottomless need for the AI industry, and that a quiet market in workplace records has grown large. Practitioners say it could have as much impact on workers and society as the data centers in the news. Ansari expects selling workflow data to become more common than ads as a revenue line for companies. The Spirit Airlines sale, where 80,000 email accounts and 500 million internal messages were offered, is the rare public glimpse of how this works.

Who it affects

Employees whose emails, chats and time-card records sit in company archives, as in the Spirit Airlines case and its flight attendants. Companies, including media firms like Wired, that receive offers for their old records. Freelance writers and graduate students, whose work Handshake offers to buy, as in the $1,000 offer for a single story. Startups that are winding down and are courted by brokers such as Sunset. The data startups themselves, which Ansari says are now a large share of new YC companies: 40 of 65 in the latest batch by the author's count.

How to use it

This is reporting, not a product, and nothing here is a how-to. What a reader can take from it is what the offers look like. Brokers pitch by cold email with large one-time fees, or a modest single-story payment like Handshake's $1,000. Sunset says it scrubs and anonymizes the data before brokering the deal. The author's own stance is a useful marker: he is not authorized to sell Wired's editorial documentation and did not offer it to Tenkara.

How solid is it

The reporting rests on first-hand pitches, a Handshake representative who confirmed the offer, and interviews with Foody and Ansari. The big numbers are weak: the $350K to $1.8M, $2 million and $45B figures come from cold emails, and the author could not verify Ansari's $40 to $50 billion a year. The author does correct Ansari on YC (40 of 65, not about half). Mercor's claim of partnerships with nearly all major labs is the company's own. Anthropic denies buying from live companies and OpenAI did not respond. The source does not say whether any lab bought Wired's records. The piece is a newsletter essay with a clear point of view, and the author's own concerns appear in his voice.

Risks and caveats

Privacy is the main risk. Industry players say technology has solved it, and Foody says de-identification can be verified, but Tobin-Miyaji of EPIC says it is not clear anonymizing can fully work, since unstructured emails and Teams messages may include medical or family matters. Ritter says norms for data acquisition do not exist yet. Workers may be enlisted, not always with explicit consent, in building systems meant to replace or automate their work. The Spirit Airlines sale is on hold and the source gives no dates or outcome. Ansari's forecast of a trillion-dollar market is a prediction from a company head with a stake in it.

“I think it’s going to become more common than ads as a revenue line item for companies”

— Ali Ansari, Micro1 CEO