Forbes: The Firms Paying You Are Also Selling to Chinese Labs

If you do AI training work, it is reasonable to want to know where your work ends up. Forbes reported in August that the answer is more complicated than most platforms say.

Short answer: Yes, Forbes reports that several US data-labelling firms sell training datasets to Chinese AI labs alongside their American clients. It does not affect whether you get paid, but it is worth knowing whose model your work may end up in.

What was reported

The reporting names specifics rather than gesturing at a trend. AfterQuery at $50 million or more in recurring revenue from Chinese labs. Mercor at roughly 2% of Q2 revenue, with an engagement manager hired specifically for China. Surge AI actively courting the region. Turing working with ByteDance according to internal documentation Forbes reviewed. Aggregate spending from the top six Chinese labs estimated at $500 million a year.

What the listings pay

None of this changes whether you get paid, and we are not going to pretend it does. Across 582 live listings on our board that publish a rate, the median top-of-range is $90 an hour, reaching $300. Payment terms are unaffected by who the end customer is.

Source: 582 live listings on this board that publish a rate, read directly from each posting on 2026-09-06. Listings without a published rate are excluded rather than estimated.

How this compares across the board

A rate only means something next to the alternatives. This is every category we track with at least five listings publishing a rate, ranked by median top-of-range, so you can see where this work sits rather than taking a single number on trust.

Same source and date as above. Categories are matched on listing title, so a role can appear in more than one.

What it means for you

What it changes is your ability to make an informed choice. Expert-annotated data is described in the reporting as a parallel and largely unregulated lever, meaning there is no export-control regime governing it the way there is for chips. If where your work ends up matters to you, that is a decision you should get to make deliberately rather than discover later.

Why this is not covered by export controls

Semiconductor export restrictions have dominated the policy conversation for three years. Expert-annotated training data has had almost none, despite being a direct input to model capability.

The Forbes framing is that the same vendors who built the pipelines pushing US frontier labs forward are selling comparable pipelines into Beijing, and that this operates as a parallel and largely unregulated lever. Whether that changes is a policy question nobody has answered yet.

For contractors, the practical consequence is that platform terms are unlikely to tell you which end customer your project serves. Most listings describe the work and the rate, not the buyer.

What you can actually control

You can ask. Some project briefs name the client, and where they do not, support will sometimes answer a direct question about whether a project serves a domestic or overseas lab.

You can also choose your projects. Work for US government and defence-adjacent clients is generally identified as such because it carries eligibility requirements, so those projects are the easiest to verify.

And you can hold two things at once: this is legitimate paid work with real payouts, and you are entitled to know roughly where it goes. Those are not in conflict.

Who should apply

Two checks before you spend time on an application. Confirm the role accepts applicants from your country with the eligibility checker, since a meaningful share of listings carry location requirements. Then run the advertised rate through the take-home calculator, because this is contract work and the headline figure is before self-employment tax.

Applications complete on the hiring platform and usually take a few minutes, with a short skills assessment in place of an interview. Fill in every credential, language and professional background field on your profile. Those are what route you to the better paid listings, and most applicants leave them blank.

Frequently asked questions

Do US data firms really sell to Chinese AI labs?

Forbes reported in August 2026 that several do, estimating the top six Chinese AI labs spend roughly $500 million a year with American data labellers.

Which companies were named?

Forbes named AfterQuery, Mercor and Surge AI, and reported that internal documentation suggests Turing has worked with ByteDance.

Does this affect whether I get paid?

No. Payment terms are unaffected by the end customer. Across 582 live listings publishing a rate, the median top-of-range is $90 an hour.

Is this legal?

The reporting describes expert-annotated training data as a parallel and largely unregulated lever, meaning there is no export-control regime covering it as there is for semiconductors.

Can I find out who my work is for?

Sometimes. Some briefs name the client, and support will occasionally answer directly. Government and defence-adjacent work is usually identified because it carries eligibility requirements.

Should I stop doing this work?

That is a personal call and we are not going to make it for you. The work is legitimate and paid. Knowing this lets you decide deliberately rather than find out afterwards.

How much of Mercor's revenue comes from Chinese labs?

Forbes reported roughly 2% of Q2 revenue, off an annualised run rate that crossed $2 billion in June, and noted Mercor had hired an engagement manager for China.

Sources

1. Forbes, Silicon Valley's other China problem: it's training their AI
2. Crypto Briefing, American data companies earn $500M a year from Chinese AI labs
3. AI Weekly, Surge AI and Mercor reportedly sell training data to Chinese labs

Keep reading

Browse every live role in this categoryPay Index: real rates from live listingsxAI Cut 500 Generalist Tutors and Promised to 10x Its SpecialistsxAI Is Hiring Thousands of AI Tutors at Up to $65 an Hour