In 2026, doctors and lawyers are being hired to train artificial intelligence, sometimes at striking rates: up to $180 an hour for US doctors on Outlier, according to Rest of World, and up to $400 for certain specialist lawyers, according to Bloomberg Law. Behind these figures, skilled work is taking a form familiar to delivery riders and online gig workers: fragmented assignments distributed and paid for by platforms that set the rules.

It was assumed that platform-based gig work would struggle to reach highly qualified professions. AI model training suggests almost the opposite: the rarer the expertise, the more attractive it becomes to split it into separately purchasable services. The danger is not necessarily that doctors or lawyers disappear, but that some of their expertise moves from a profession into a catalogue of tasks.

Why are AI companies hiring doctors and lawyers?

Models have absorbed vast quantities of online text. To move beyond this general material, developers need specialist answers, difficult reasoning and assessments that distinguish plausible solutions from genuinely sound ones. This is the role of human work in post-training.

A doctor can devise a complex clinical case, review a model's answer and identify weaknesses. A lawyer can compare arguments, spot questionable interpretations or rank responses. A physicist or engineer can check a result's consistency. Rather than treating a patient or representing a client, the professional produces a unit of judgment to improve a system.

According to Rest of World, Scale AI's Outlier offers US doctors up to $180 an hour for complex cardiac scenarios. Bloomberg Law reports up to $400 an hour for civil rights lawyers; other listings mention $170 for condensed-matter physicists and $120 for software engineers. These ceilings are neither guaranteed wages nor benchmarks for France.

What is sold is not a job but a narrowly defined contribution.

How does this resemble the delivery and online gig economy?

The similarity is not pay but work structure. A platform aggregates scattered demand, turns it into standardized assignments, allocates tasks and controls access to future work. Workers provide skills; the intermediary manages the commercial relationship and can change its terms.

In delivery, a journey becomes a standalone service. In AI training, a reasoned diagnosis, legal review or scientific check becomes the saleable unit. A degree does not remove dependence on the platform; it merely changes what some tasks might command.

Rest of World also reports geographic differences on Outlier: up to $7.50 an hour for Tagalog work in the Philippines and $12.50 for Swahili in Tanzania. That does not make a US cardiac scenario equivalent to a language task. Rather, the examples show how specialization, location and pricing power combine.

Publicité

Does $400 an hour really protect against insecure work?

A high rate pays for an hour when work is available. It says nothing about available hours, consistent orders, time spent qualifying or bargaining power. Comparing it with a stable monthly salary means comparing different realities.

Platforms are not mere job boards. They select, allocate and charge for access to expertise. According to AInvest, intermediaries such as Mercor retain roughly 30–40% of gross payments. This describes a reported business model, not a way to calculate every expert's take-home pay.

The deeper question is who owns the output. MIT researcher Danielle Li notes that a company training a model retains a form of an expert's work after an assignment ends. In traditional consulting, the client receives advice or a service. Here, professional judgment can also improve a reusable tool that may later perform some of the same tasks.

That is the paradox: expertise is valuable enough to earn high rates today, yet some of its value is captured by a system the expert does not control. This does not mean every contribution automatically replaces its author. It means a short contract and its lasting effects no longer share the same timescale.

Is this already happening in France?

Caution is needed. Available evidence does not establish that French doctors or lawyers already perform platform tasks at scale on Outlier or Mercor, or receive US rates. Calling it an established French market would go beyond the facts.

France is nevertheless affected by another aspect: breaking knowledge-based professions into automatable tasks. Figures attributed to Coface and the Observatory of Threatened and Emerging Jobs, cited by Vosposts, put exposure to automation at around 27% for legal, financial and creative professions and 29% for engineering. Exposure does not mean the same share of jobs will disappear.

Lawyers reviewing contracts and bank customer advisers illustrate the shift. A full profession involves relationships, responsibility and decisions, but certain document processes can be separated. Once isolated, they can be automated, outsourced or used to evaluate a model. These are distinct economic options, not inevitable stages.

The French question is therefore not just how many jobs AI will destroy, but who will determine the value of the remaining human tasks. A profession may keep its title, qualifications and prestige while losing some control over how its work is organized and negotiated.

Is professional judgment the next frontier of platform work?

Task-based work was long associated with activities thought easy to replace. AI training reverses that assumption: it buys precisely what machines cannot yet do well. Human intelligence has not become useless; it is converted into a measurable service that clients can activate and discontinue.

The key divide is no longer simply manual versus graduate work. It separates those who control client relationships, working terms and future uses of their output from those who supply fragments to infrastructure owned by others. AI may be shifting this boundary fastest.

Frequently asked questions

How are doctors and lawyers paid to train AI?

Specialist platforms hire professionals to write complex cases, annotate answers and verify model outputs. Outlier and Mercor are among the intermediaries mentioned. Assignments are temporary, with terms set by platforms.

Are French doctors and lawyers already affected?

Available information does not establish a comparable large-scale French gig market for these professionals. Exposure to automation in French legal, financial and engineering work is documented, however. This transformation risk must be distinguished from actual platform recruitment.

Why can highly qualified professionals become gig workers?

A platform can buy a specific slice of expertise without providing ongoing employment. It controls assignments and conditions, while experts depend on available work. Qualifications alone do not guarantee economic independence.

Are the advertised rates guaranteed?

No. Rates of up to $180 an hour for some doctors and $400 for some lawyers are reported US market ceilings, not guaranteed regular income. They also cannot be used to estimate French rates.

Sources