Weekly market news · 10 Oct to 16 Oct 2026
The week in AI training work: micro1 passes a $1B run rate, Labelbox raises $100M, open roles hold steady
Week of 10 to 16 Oct 2026: open roles held steady on the platforms read the same way all week, 12 platforms joined our board, micro1 reported a $1B run rate and Labelbox raised $100M.
AITraining.jobs research team · Published
On the 12 platforms we read the same way all week, open AI training roles ended exactly where they began: 1,552 on 10 Oct and 1,552 at the end of 16 Oct 2026, as 140 new listings replaced 140 removals. The board as a whole grew from 2,547 to 3,635 open roles, but almost all of that is how we read it, not the market: 12 platforms' own job boards were read for the first time, and the copies of other platforms' roles that OpenTrain AI re-lists were dropped.
Outside our data, it was a week of money and new entrants: micro1 said it passed a $1 billion annual run rate, Labelbox (the company behind the Alignerr expert network) raised more than $100 million, and Toptal and GLG both launched expert services for AI labs.
- roles open on the board at the end of 16 Oct 2026
- 3,635
- change in open roles on the 12 platforms read the same way all week
- 0%
- new listings on those platforms
- 140
- platforms whose own job board was read for the first time
- 12
Where the open roles moved
Law, Medicine & Finance remained the largest kind of AI training work, with 1,074 open roles on the board at the end of 16 Oct. On the comparable platforms it grew from 348 to 369 (+6%), with 55 new listings. Transcription, audio and speech rose from 236 to 250, and multilingual work held at 551 (from 557).
General evaluation and labeling work shrank: AI evaluation and rating (RLHF) fell from 65 to 56 open roles (-13.8%) and data annotation from 80 to 72 (-10%). Coding and software evaluation went from 90 to 85 on the comparable platforms, while the whole board held 284 coding roles once the newly read boards are counted.
Among the comparable platforms, micro1 moved most: 70 new listings and 84 removals took it from 255 to 241 open roles. Appen grew from 46 to 55 and xAI's roles fell from 22 to 15. No country went from no open roles to some, or from some to none, on these platforms.
Pay
The median advertised hourly rate of the open roles on the comparable platforms rose from $60 to $65. That is a change in which roles were open, not a raise: within the larger kinds the medians barely moved. Multilingual work stayed at $36.50, transcription and audio at $30 and coding at $65. Law, Medicine & Finance eased from $75 to $72.50, RLHF evaluation rose from $45 to $47.75 on fewer open roles, and data annotation fell from $32 to $28.
Handshake AI started to show its fellows a pay dashboard this week, which itemizes each pay week by base pay, incentives, adjustments and disputes (see the platform news below).
New on our board this week
12 platforms' own job boards were read for the first time this week: SuperAnnotate (SME) with 377 open roles, Turing with 194, Anyone AI with 128, Perle Labs, Claru, Babel Audio, RemoExperts, OneForma, Sovrano AI, Vetto, Pasiflora AI and Pila8. Being new to our count does not make them new to the market, which is why they stay out of every rise and fall above.
Three boards were read more fully: Alignerr (328 roles added on 14 Oct), Handshake AI (159 on 14 Oct) and Mindrift (158 on 15 Oct). And on 14 Oct we stopped counting OpenTrain AI's copies of roles that other platforms post themselves (666 removed that day): OpenTrain re-lists other platforms' openings, and each copy counted one role twice. Its own roles are still listed, and the originals are listed under the platforms that post them. Mercor, which took down 72 listings in one day on 12 Oct, is left out of the comparison as well.
What the AI labs are hiring for
On 17 Oct 2026, 16 of the 25 AI labs whose public job boards we read had 101 roles open on the teams that buy, run and evaluate human data, 2.9% of their 3,505 open roles. xAI had the most, 24 of its 304.
The most common of these jobs is the one closest to AI training work itself: AI tutors and raters, 25 roles at 4 labs. Robot data collection (teleoperators, robot pilots, data creators) had 14 roles at 1X, Figure AI, Physical Intelligence and Skild AI, and 7 roles buy or license data, at Mistral AI, OpenAI and Thinking Machines Lab. That last line connects to the week's news: AI training vendors are now bidding for companies' own data, too.
Platform news
What was reported about the platforms this week, each item with the outlet that reported it.
· micro1
micro1 says it passed a $1 billion run rate and priced its Series B at $4 billion
micro1 announced that it has passed $1 billion in annual run rate while profitable, priced its Series B at a $4 billion valuation and ran a tender offer of about $25 million for early employees. The figures are the company's own and unaudited.
Sources: micro1 release via AiThority, 13 Oct 2026; Dealroom, 13 Oct 2026
· Alignerr
Labelbox, the company behind Alignerr, raises more than $100 million
Labelbox announced more than $100 million in Series E financing, equity and a credit facility, at a $1.25 billion pre-money valuation, led by Infinitum Partners. Alignerr is Labelbox's network of experts who write and grade the tasks models train on.
Sources: Labelbox blog, 14 Oct 2026; PR Newswire, 14 Oct 2026
· Mercor
Mercor cut at least 50 staff, a local outlet reports; Mercor says it was not a layoff
Gazetteer SF reported that Mercor let go of at least 50 full-time employees across several teams the week before (its staff, not the contractors who do AI training work through it). A Mercor spokesperson said less than 5% of the company left, and CEO Brendan Foody said it was not a layoff. Only one outlet has reported it so far.
Source: Gazetteer SF, 14 Oct 2026
Handshake AI starts rolling out a pay dashboard for fellows
Handshake AI began rolling out a pay dashboard on 14 Oct for fellows on select projects, on desktop. It shows each pay week and project with base pay, incentives, adjustments (each with a reason) and disputes, and access expands in phases over the following weeks.
Source: Handshake AI Help Center, 14 Oct 2026
Toptal launches ExpertCrowd, an AI training service built on its expert network
Freelance network Toptal launched ExpertCrowd for RLHF, reinforcement-learning data, data labeling and model evaluation. It says the service draws on more than 30,000 domain experts in 140+ countries and earned over $5 million in recurring revenue before launch, from clients including Scale AI, Sterling Labs and Vetto.
Sources: PR Newswire (Toptal), 13 Oct 2026; TNGlobal, 14 Oct 2026
Expert network GLG launches GLG Frontier for AI labs
GLG launched GLG Frontier, a practice that supplies AI labs and model developers with vetted specialists to design evaluations, build benchmarks, and fine-tune and red-team models. It is led by Grady Gardner, who previously built AI training-data programs at Braintrust.
Source: PR Newswire (GLG), 15 Oct 2026
Scale AI and Mercor both post-trained NVIDIA's new open model
At the launch of NVIDIA's Nemotron 3.5 Super VL, Scale AI said it had post-trained the model ahead of release with enterprises such as Skydance and EY and launched a post-training service, and Mercor said it post-trained it for legal, software and accounting work with tasks built by experienced practitioners.
Sources: Scale AI blog, 15 Oct 2026; Mercor blog, 15 Oct 2026
Hearing on the sale of Spirit Airlines' company data, with Mercor and micro1 bidding, moves to 6 Nov
In Spirit Airlines' bankruptcy case, the court hearing on selling Spirit's de-identified company data to Google was taken off the 14 Oct agenda and adjourned to 6 Nov 2026. Mercor is the alternate bidder, and micro1 has filed notice of a competing bid: AI training vendors are now bidding for companies' own records.
Sources: U.S. Bankruptcy Court S.D.N.Y., amended hearing agenda, 12 Oct 2026; Same case, consumer privacy ombudsman report, 5 Oct 2026
Bloomberg: AI training jobs are attracting doctors, lawyers and engineers
Bloomberg ran a feature under that headline on 16 Oct; we could read only the headline, as the article is paywalled. On our board, Law, Medicine & Finance is the largest kind of AI training work, with 1,074 open roles at the end of 16 Oct.
Source: Bloomberg, 16 Oct 2026
· micro1
Earlier this month: micro1 commits $1 billion to buying company data
micro1 said it will spend $1 billion over the next 12 months buying and licensing operational data from companies (procedures, knowledge bases, CRM records) to build training environments, financed by Citi and Hercules Capital.
Sources: The Next Web, 9 Oct 2026; Pulse 2.0, 11 Oct 2026
The week in our data
10 Oct 2026 to 16 Oct 2026, read from the job boards of the 29 platforms with open roles on AITraining.jobs. Each count opens the live list, which keeps moving after this week.
Open roles by kind of work
The board's count at the end of 16 Oct 2026, then the same kind on the 12 platforms read the same way all week, from the start of 10 Oct 2026 to the end of 16 Oct 2026. Median pay is the middle advertised hourly rate, shown where at least 20 roles state one at both ends.
| Kind of work | On the board | Comparable, start to end | New listings | Median pay, start to end |
|---|---|---|---|---|
| Law, Medicine & Finance | 1,074 | 348 to 369(+6%) | 55 | $75 to $72.50 |
| Multilingual & Localization | 1,002 | 557 to 551(-1.1%) | 22 | $36.50 to $36.50 |
| Transcription, Audio & Speech | 519 | 236 to 250(+5.9%) | 31 | $30 to $30 |
| Coding & Software Evaluation | 284 | 90 to 85(-5.6%) | 9 | $65 to $65 |
| Math & STEM | 226 | 114 to 113(-0.9%) | 4 | $65 to $65 |
| AI Evaluation & Rating (RLHF) | 147 | 65 to 56(-13.8%) | 8 | $45 to $47.75 |
| Data Annotation & Labeling | 134 | 80 to 72(-10%) | 8 | $32 to $28 |
| Robotics & Physical-AI Data | 65 | 17 to 12 | 0 | too few rates |
| Agentic Tasks & RL Environments | 62 | 9 to 10 | 2 | too few rates |
| Red-Teaming & AI Safety | 52 | 1 to 1 | 0 | too few rates |
| Writing & Content | 36 | 16 to 16 | 1 | too few rates |
| Research Studies & Surveys | 34 | 19 to 17 | 0 | too few rates |
What changed in what the board reads
These platforms are left out of every "rose" or "fell" in this post: their counts moved because of how they are read, not because the market moved.
- SuperAnnotate (SME): its own job board read for the first time on 10 Oct 2026, 377 roles open at the week's end.
- Turing: its own job board read for the first time on 10 Oct 2026, 194 roles open at the week's end.
- Anyone AI: its own job board read for the first time on 11 Oct 2026, 128 roles open at the week's end.
- Perle Labs: its own job board read for the first time on 10 Oct 2026, 67 roles open at the week's end.
- Claru: its own job board read for the first time on 14 Oct 2026, 60 roles open at the week's end.
- Babel Audio: its own job board read for the first time on 11 Oct 2026, 43 roles open at the week's end.
- RemoExperts: its own job board read for the first time on 10 Oct 2026, 36 roles open at the week's end.
- OneForma: its own job board read for the first time on 14 Oct 2026, 24 roles open at the week's end.
- Sovrano AI: its own job board read for the first time on 11 Oct 2026, 20 roles open at the week's end.
- Vetto: its own job board read for the first time on 14 Oct 2026, 20 roles open at the week's end.
- Pasiflora AI: its own job board read for the first time on 14 Oct 2026, 5 roles open at the week's end.
- Pila8: its own job board read for the first time on 11 Oct 2026, 3 roles open at the week's end.
- OpenTrain AI: a batch left out on 14 Oct 2026, 666 roles removed in one day.
- Alignerr: read more fully from 14 Oct 2026, 328 roles added in one day.
- Handshake AI: read more fully from 14 Oct 2026, 159 roles added in one day.
- Mindrift: read more fully from 15 Oct 2026, 158 roles added in one day.
- Mercor: a batch left out on 12 Oct 2026, 72 roles removed in one day.
Biggest moves on the comparable platforms
- micro1: 255 to 241 open roles, 70 new listings and 84 removed.
- Appen: 46 to 55 open roles, 16 new listings and 7 removed.
- TransPerfect DataForce: 59 to 66 open roles, 10 new listings and 3 removed.
- xAI (SpaceXAI): 22 to 15 open roles, no new listings and 7 removed.
- Braintrust: 82 to 78 open roles, 6 new listings and 10 removed.
- Welo Data: 232 to 235 open roles, 23 new listings and 20 removed.
Countries opened or closed
No country went from no open roles to some, or from some to none, on the comparable platforms this week.
How this post is counted
- The week is the seven full days (UTC) before the day of publication. A role is open at a moment when the platform's board lists it then; a new listing is one first seen during the week; a removal is one no longer listed.
- Every "rose" or "fell" is measured on the platforms read the same way all week. A platform is left out when its own board was read for the first time, or when on one day at least 25 of its roles appeared or disappeared at once and they made up at least 30% of what it had open that morning: that is how a change in what the board reads looks, not how hiring looks.
- Pay is the advertised hourly rate (the middle of a stated range); a median is given only where at least 20 roles state one. A country counts as opened or closed only when at least 3 roles were open to it on the other side of the week.
- Platform news comes from the outlet or company named under each item, linked and dated. The post's words are fixed on its day; a correction is added below them with its date.
Cite and follow
AITraining.jobs (2026). The week in AI training work: micro1 passes a $1B run rate, Labelbox raises $100M, open roles hold steady. https://aitraining.jobs/news/2026-10-17
Free to quote with credit under CC BY 4.0. The monthly view of the same market is in the monthly data stories, every figure in the research reports.