Mercor’s cofounders last month became billionaires after raising $350 million on a $10 billion for the AI training startup. Just over a week later they allegedly canned an AI project that involved thousands of contractors. Hours later it offered to rehire workers on a lower hourly rate, according to several people impacted.
Contractors who worked with startup Mercor say they found themselves locked out of Slack on Tuesday after the AI training project they worked on was abruptly scrapped.
The San Francisco startup, which was valued at $10 billion last month, allegedly emailed contractors on Wednesday to inform them that the project code named Musen had been cancelled.
Forbes spoke with five Mercor contractors who worked on the project that was focused on reviewing video and audio from Meta’s Facebook and Instagram short-video platform Reels. The contractors were part of a Slack group that shared updates on the project that at one point included over 5,000 people, according to these contractors.
Contractors told Forbes that they had been working on the project for several months and Mercor managers had told them in October that the project was expected to run until at least December. Mercor did not officially identify the client behind the project but workers were told it was Meta and described exclusively working with content from Meta’s social media apps to train AIs to identify people and products in short form videos.
“It was all very sudden. I have never seen anything like that and have worked on a few AI projects,” one Mercor contractor told Forbes. All of the contractors interviewed requested not to be named or identified.
Mercor had been paying workers on the project $21 per hour but contractors complained that in the last few weeks the project had at times been paused and promised hours were cut back. “It felt like a slap in the face. It’s very disrespectful to ask us to do more work for less money,” said another contractor.
After the Slack group was shut down and former contractors took to Reddit to regroup and work out what happened, Mercor emailed them around 1pm ET: “We’ve carefully reviewed contributor feedback regarding task availability, hour caps, and workload consistency. We’ve heard you – and we’re committed to creating a more stable and predictable environment for everyone moving forward.”
The email also included an offer for some workers to take on new roles on a project dubbed Nova. Mercor promised “steadier task volumes” and more hours per week of work but also set a new hourly rate of $16 per hour. That’s lower than the state minimum wage set in locations like California, Washington and Connecticut.
“While this reflects a change from the current structure, our goal is to offer greater earning stability and consistent access to work, rather than fluctuating opportunities,” Mercor said in the email to its contractors that was seen by Forbes.
“It’s contract work but we are real people who deserve some notice, or warning or some consideration,” another former Mercor contractor told Forbes. “I know we are working with AI but we don’t work for AI. You don’t just dump thousands of people, that’s not just right.”
Mercor’s head of head communications Heidi Hagberg said: “It’s inaccurate and we decline to comment further.”
Mercor’s move to slash wages for some of its contractors comes just weeks after a new $350 million funding round boosted the company’s valuation to $10 billion, up from $2 billion just months earlier, and crowned its three 22-year-old cofounders as the world’s youngest self-made billionaires.
Bay Area high school friends Adarsh Hiremath, Brendan Foody and Surya Midha dropped out of college to found Mercor, originally billed as an AI recruiting platform for software engineers.
Got a tip? Contact reporter Iain Martin at [email protected] or +1 (646) 739-6427 on Signal or WhatsApp.
In the process, the trio stumbled into the booming area of data labelling for AI labs. Mercor said it had hit $500 million in annualized revenue run rate in September shortly after debuting on the Forbes Cloud 100 list of the top private cloud computing companies. (It would not disclose its trailing revenues.) That’s largely thanks to a huge shakeup in the data labelling industry that had already coined several three comma fortunes for the founders of rival data labelling labs like Scale and Surge.
Meta announced in June that it was buying 49% of one of the sector’s largest players Scale for $14 billion in a move to hire its cofounder and CEO Alexandr Wang, who is now Meta’s chief AI officer. That move sparked a feeding frenzy from companies like Mercor, Surge, and Turing that bet that rival AI labs would not want to work with a company now partially owned by a major rival.
Mercor’s CEO Brendan Foody said in September that its revenue had grown fivefold since March but the boom has also sparked tensions. That same month, Scale sued Mercor alleging that the startup had stolen trade secrets. “It’s not something we spend a lot of time thinking about,” Foody told Forbes when asked about the complaint.
Big tech’s massive investment in generative artificial intelligence has fueled demand for thousands of humans to work often short-term and flexible jobs to label training data, coach nascent chatbots, and review outputs for safety. In the early days, this often involved low-paid “clickworkers” in Africa and South East Asia but as models have advanced so have the demands on workers with many labs hiring specialists and masters-level graduates to work on specialist topics.
Mercor might only be offering data annotation workers on Project Nova $16 per hour but it is also running job ads for lawyers, journalists and doctors who it says it is willing to pay up to $200 per hour to answer questions about their work to shape new AI models.
Several of the Mercor contractors with whom Forbes spoke said that they had already signed up for the new project at the lower rate. “I’m part time but for a lot of my colleagues this was their bread and butter and they have kids.“
