Doctors and Lawyers Now Train AI for Up to $400 an Hour

Bloomberg reported on 16 October 2026 that the work of training AI models is drawing white-collar specialists at far higher rates than the generalists who built the field. It says recruiters are searching for software engineers at up to $120 per hour, condensed matter experts at $170 per hour and civil rights attorneys at $400 per hour, all to “help train next-generation AI systems.”

Average pay at Mercor, a three-year-old platform that recruits white-collar data workers, is now $110 per hour, Bloomberg reports, though the company said it did not have data on what percentage of tasks pay that rate.

Business Engineer Video · DATA LABELING

A Business Engineer explainer on data labeling, the human work behind an AI model. It covers the concept in general and says nothing about Bloomberg’s report.

What Bloomberg Reported

The report, by Morgan Meaker, says data work has long been associated with low wages and is now attracting specialists who can command much higher sums. It says LinkedIn named data annotators as the fourth-fastest-growing job role in the US this year.

Bloomberg describes two US data workers: Brook Hansen, who started data work after losing a customer service job and earns a couple of hundred dollars a week, and Anu Brixey, a radiologist, who it says makes that same amount in an hour.

What Was Posted on X

Bloomberg’s own post on 16 October promoting the article this piece is based on. It says the gig workers training AI increasingly have medical degrees, law degrees and specialized expertise, and that some command hundreds of dollars an hour.

The Pay Ladder

In the industry these specialists are known as “experts,” Bloomberg says, and they share the market with lower-paid generalists doing less specialized tasks. British platform Prolific declined to say how many of its tasks are completed by its experts, who make up 30% of its total participants, the report says.

Ana Price, a vice president at Prolific, told Bloomberg that platforms are paying more to attract busy professionals: “Given the need for white collar — these are typically people who I would categorize as time-poor but more money-rich,” she said. “So the compensation needs to be compelling.”

Brixey told Bloomberg she signs up for at most two projects a year, motivated less by the pay than by the technology’s potential: “I’m doing it for improved patient care.”

Where the Platforms Stand

Bloomberg says Amazon’s Mechanical Turk, once the most important platform for American data workers, shut down last month after 21 years. Mercor, it says, has been discussing new funding at a valuation of $20 billion, and Micro1 is valued at $4 billion.

Micro1 founder Ali Ansari wrote in January, as quoted by Bloomberg, that “Every hour of work can simultaneously: Run the organization, train AI models, generate additional revenue for the organization.” Bloomberg also reports that Meta started installing software in April to track employees’ clicks, keystrokes and mouse movements, citing a memo: “This is where all Meta employees can help our models get better simply by doing their daily work.”

We have covered Micro1’s $4 billion Series B pricing and its commitment of $1 billion to acquire company data, and Mercor’s earlier $20 billion valuation.

The Generalists and the Skeptics

Lower-paid generalists say conditions are deteriorating, with harder tasks and steady work harder to find, Bloomberg reports. Hansen, an organizer for Turkopticon, said platforms now often want workers who specialize in subjects like math or biology.

Antonio Casilli, a professor at the Polytechnic Institute of Paris, told Bloomberg that highly paid experts take part in relatively few projects compared with generalists, and that such experts “are just the cherry on top.”

MIT professor Danielle Li told Bloomberg a small group of data workers could in theory replace or substantially devalue an entire profession: “When you are supplying data, other people’s choices are very much impacted by yours.” On 17 October we reported Anthropic’s report on falling junior technical hiring, and on 16 October its survey of staff expectations for AI.

The Structural Read

As we read it, the scarce input has moved from volume to judgment: Bloomberg says companies are now racing to secure bespoke, privately held data, after publicly available information scraped from the internet formed the foundation of today’s models.

As we read it, the platforms are pricing that scarcity: Bloomberg reports Mercor’s average pay at $110 an hour, and Prolific’s Ana Price says compensation for time-poor professionals needs to be compelling.

As we read it, employers are becoming a supply too: Bloomberg cites Micro1’s Ali Ansari arguing that every hour of work can also train AI models, and a Meta memo on using employees’ daily work to improve its models.

Three Implications

DATA-WORK PLATFORMS Bloomberg reports that Prolific’s experts make up 30% of its participants and that Mercor has been discussing funding at a $20 billion valuation, while Mechanical Turk shut down last month after 21 years.

PROFESSIONALS WEIGHING THE WORK Bloomberg quotes radiologist Pauley Chea saying “People are worried,” and saying the uncertainty is particularly pronounced among medical students and residents.

EMPLOYERS Bloomberg reports that Meta started tracking employees’ clicks, keystrokes and mouse movements in April to help its models, and quotes Micro1’s founder arguing that employees’ hours can also train AI models.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Four Intelligence Moats.

The framework names four moats, one per data paradigm, and describes the Verifier Moat as “built by RLVR on domain-specific reward signals, building into a new asset class for reasoning-heavy verticals.”

As we read it, the specialists in Bloomberg’s report, such as radiologists checking scans and attorneys reviewing answers, are the people who supply those domain-specific signals. The framework predates the report and does not discuss it.

What Is Not Established

Bloomberg’s hourly rates for engineers, condensed matter experts and civil rights attorneys come from recruiter postings it describes, not from payroll data. Mercor did not say what share of tasks pays its $110 average, and Prolific did not say how many tasks its experts complete.

We read Bloomberg’s article in full on 17 October 2026. We did not contact Mercor, Prolific, Micro1 or Meta.

The Bottom Line

Bloomberg reported on 16 October 2026 that recruiters are offering up to $400 an hour for specialists such as civil rights attorneys to help train AI, and that Mercor’s average pay is $110 an hour.

The report places those rates against generalist data workers who say their conditions are getting worse, and quotes researchers who question how much of the market the highly paid experts really represent.