Scale AI Remote Tasks in 2026: Where the Work Went

By Mufy Pachorawala, founder of getAIwork · Last updated: September 10, 2026

What actually happened to Scale AI?

Three things in fifteen months, and they explain most of what people notice about the task supply.

On June 12, 2025 Scale announced what it called the next phase of the company’s evolution. Meta invested, founder Alexandr Wang left to work on Meta’s AI efforts while remaining a board director, and chief strategy officer Jason Droege became interim chief executive. Scale’s own wording described Meta as holding a minority of Scale’s outstanding equity at a valuation of over $29 billion. Press reporting put concrete numbers on it: roughly $14.3 billion for a 49% non-controlling stake. Scale has never stated those figures itself, which is itself informative.

Then the customers left. Within two days of the announcement, Google, reported to be Scale’s largest customer, was planning to split. OpenAI wound down its work by June 18. Microsoft ended its partnership the same month. xAI was reported to be exploring alternatives. If your biggest buyers are frontier labs and one of them buys half of your supplier, the others do not stay.

On July 16, 2025 Scale cut 200 employees, 14% of its staff, plus 500 contractors, with the cuts concentrated in data labelling while enterprise and government sales were expanded. Droege’s memo said the company had scaled its core data-labelling business too quickly.

And on July 30, 2026, Scale appointed Francis deSouza as chief executive, previously chief operating officer and president of security products at Google Cloud and before that chief executive of Illumina. His stated priorities are getting Scale’s solutions into more businesses and governments, delivering data for AI labs, and showing value through provable outcomes. If you are reading an article that still says Alexandr Wang runs Scale, it is at least fourteen months stale.

The direction is visible in the hiring. Of 217 open roles on Scale’s careers board on September 10, 2026, thirty-nine carry public sector, government, defence or national security in the title and forty-four are located in Washington DC. The boilerplate on every listing names Meta, Ernst and Young, Mayo Clinic, Time Inc., the Government of Qatar and US government agencies including the Army and Air Force. Scale is now an enterprise and defence contractor that also runs crowd platforms, rather than the other way round.

The three Scale worker platforms, and which is which

Almost every guide treats Scale’s crowd work as one thing. Scale’s own Terms of Use name three separate platforms with three separate support channels, and they sit at very different points on the ladder.

Remotasks deserves its own article and we have given it one. The short version for here: the site still loads, accounts still log in, and it is still covered by current terms, but its own sign-up flow now redirects blocked applicants elsewhere. Its exact wording is worth quoting, because it is the company answering the question directly: Unfortunately, we aren’t currently accepting new contributors from your location on Remotasks. However, we encourage you to check out exciting opportunities to train Generative AI on Outlier.ai.

Outlier: the route that actually works

Outlier is where Scale points people, and it is accepting sign-ups. There is a live opportunities board covering maths, law, full-stack software engineering, coding in Simplified Chinese and Spanish, agent-building projects and region-specific listings.

The ownership has changed and nobody has reported it. Outlier’s FAQ still describes it as a platform operated by Scale AI, but its Terms of Use effective May 11, 2026 name the contracting party as Scale AI, Inc., a Delaware corporation, and its affiliates, including Smart Ecosystem Inc., doing business as Outlier AI. Every page footer now reads powered by Scale AI above © 2026 Smart Ecosystems. We found no announcement or press coverage explaining the restructuring, so we are reporting the entity names as fact and leaving the reason alone.

One practical correction that will save you an evening: Outlier’s old Greenhouse job board is dead. Those URLs now return “the job board you were viewing is no longer active”. Applications go through app.outlier.ai and nowhere else. A large amount of ranking content, aggregator listings and forum advice still links the dead board.

To apply you need a valid ID and mobile phone from your country of residence, a current resume, and a LinkedIn profile showing your education and work history. Outlier says you need at least an associate degree, that some projects require a bachelor’s, master’s or PhD, and that it runs a selective screening process. Onboarding is described as taking 30 to 90 minutes. You work as an independent contractor and there is no visa sponsorship.

Outlier says it operates in 50 countries. It publishes no country list anywhere on its site, so eligibility only becomes clear per opportunity. If you want a view of what is open across the whole category rather than one platform’s board, our AI training platforms hiring now roundup is the faster check.

What do Scale’s platforms pay?

This is the section where honesty is more useful than a number. Outlier does not publish a rate. Its FAQ answers the question “how much does Outlier pay?” with: Rates vary by expertise, project complexity, and location. You’ll always see the tasking rate before starting any project.

So any article quoting you a specific Outlier hourly range is either quoting a single project someone saw, or quoting the dead Greenhouse listings. Neither is a rate card. What Outlier does publish is that it has paid over $500 million to experts, that pay is weekly, and that it offers competitive pay plus quality-based rewards.

That last row is the interesting one. Scale publishes salary bands for its employees on the same careers board where its hourly contractor role publishes nothing. That is pay transparency law working exactly as written, and it is a neat illustration of which side of the line crowd work sits on.

The weekly payment cycle is genuinely a point in Outlier’s favour. Compared with the roughly six-week cycle we describe in our Appen review, weekly on Tuesdays is a meaningful quality-of-life difference when the work is irregular.

What is the Human Frontier Collective?

The least-documented of the three, and the most interesting if you have a research background. HFC sits at hfc.scale.com, is badged Powered by Scale AI, and has its own support address at Scale. Its positioning is fellowship rather than gig work: apply to the fellowship, define what right means in your field, build evaluations and reasoning standards that frontier systems have to meet.

The testimonials name doctoral researchers from MIT, Stanford, Harvard Medical School, Johns Hopkins and UC Davis, and one member describes authoring legal benchmarks later accepted at major machine learning conferences. There is an open sign-up.

It publishes no pay figures at all. That is not necessarily a warning sign at this end of the market, where engagements are negotiated, but it does mean you cannot compare it to anything before applying. If you hold a doctorate and this work interests you, the credential-led end of the category is where rates are highest across the board, as we found in our TELUS Digital review and Surge AI review.

Scale’s own careers board has a contractor role, and most people miss it

Tucked into Scale’s corporate careers board is a part-time hourly posting called Subject Matter Expert, open in Canada and the United States, last updated in August 2026. Its wording: Part-time (Hourly). Work as many or as few hours as you’d like, with the flexibility to set your own schedule based on your availability.

The work is quality auditing rather than labelling: structured audits of task outputs, defining quality benchmarks, scoring rubrics and evaluation criteria. It explicitly excludes contributor-facing work, with no task assignment or queue management. It asks for deep domain expertise in a specialised field such as medicine, law or engineering. It publishes no rate.

If you have the credentials, this is a route into Scale that does not go through a crowd platform at all, and it is not mentioned in any of the roundups we checked.

One widely shared story that is not about Scale

A piece of housekeeping, because we keep seeing this merged. The November 2025 story about an AI company cutting thousands of contractors from a project and redeploying them at $16/hour, down $5 is about Mercor, not Scale. Several roundups have folded it into Scale’s timeline. It is a real and relevant datapoint about the category’s rate direction, and it belongs to a different company. Our Mercor review covers that platform properly.

So where should you actually apply?

The honest framing of Scale in 2026 is that it is a large, healthy, pivoting company whose crowd platforms are now a side business rather than the main event. That is neither good news nor bad news for you specifically. It just means the task supply follows Scale’s enterprise and government contracts rather than the frontier-lab boom that built the platform, and that is a slower, lumpier rhythm.

As always, do not wait on one queue. Outlier alternatives covers the platforms most people run alongside it, and AI training jobs with no experience is the right starting point if you are new to this work.

Mufy Pachorawala · Founder, getAIwork

My AI scans thousands of AI-job posts (40,786 screened so far) and I personally approve every listing before it reaches the board. I write these articles by the same rules: pay quoted only as listed, and no income promises. Read our editorial rules.

Related: Outlier alternatives · Surge AI review · TELUS Digital review · Appen review · Mercor review · AI training platforms hiring now