It’s handiest been since June that Meta invested $14.3 billion within the information dealer Scale AI, bringing on CEO Alexandr Wang and a number of other of the startup’s most sensible executives to run Meta Superintelligence Labs (MSL). On the other hand, the connection between the 2 corporations is already appearing indicators of fraying.
A minimum of some of the executives Wang introduced over to assist run MSL — Scale AI’s former Senior Vice President of GenAI Product and Operations, Ruben Mayer — has departed Meta after simply two months with the corporate, two other people accustomed to the subject instructed TechCrunch.
Mayer spent kind of 5 years with Scale AI throughout two stints. In his little while at Meta, Mayer oversaw AI information operations groups and reported to Wang, however wasn’t tapped to sign up for the corporate’s TBD labs — the core unit tasked with construction AI superintelligence, the place most sensible AI researchers from OpenAI have landed.
Mayer didn’t reply to 2 separate requests for remark from TechCrunch.
Additional, TBD Labs is operating with third-party information distributors as opposed to Scale AI to coach its upcoming AI fashions, consistent with 5 other people accustomed to the subject. The ones third-party distributors come with Mercor and Surge, two of Scale AI’s biggest competition, the folks stated.
Whilst AI labs often paintings with a number of information distributors – Meta has been running with Mercor and Surge since earlier than TBD Labs was once spun up – it’s uncommon for an AI lab to take a position so closely in a single information dealer. That makes this case particularly notable: even with Meta’s multi-billion-dollar funding, a number of resources stated that researchers in TBD Labs see Scale AI’s information as low high quality and feature expressed a choice to paintings with Surge and Mercor.
Scale AI to begin with constructed its industry on a crowdsourcing style that used a big, cheap personnel to care for easy information annotation duties. However as AI fashions have grown extra refined, they now require highly-skilled area professionals—corresponding to medical doctors, legal professionals, and scientists—to generate and refine the high quality information had to reinforce their efficiency.
Techcrunch match
San Francisco
|
October 27-29, 2025
Even supposing Scale AI has moved to draw those material professionals with its Outlier platform, competition like Surge AI and Mercor were rising temporarily as a result of their industry fashions had been constructed on a basis of high-paid skill from the outset.
A Meta spokesperson disputed the truth that there are high quality problems with Scale AI’s product. Surge and Mercor declined to remark. Requested about Meta’s deepening reliance on competing information suppliers, a Scale AI spokesperson directed TechCrunch to its preliminary announcement of Meta’s funding within the startup, which cites a diffusion of the firms’ industrial dating.
Meta’s offers with third-party information distributors most probably imply the corporate isn’t striking all its eggs in Scale AI, even after making an investment billions within the startup. The similar can’t be stated for Scale AI, alternatively. In a while after Meta introduced its large funding with Scale AI, OpenAI and Google stated they might forestall running with the knowledge supplier.
In a while after dropping the ones consumers, Scale AI laid off 200 workers in its information labeling industry in July, with the corporate’s new CEO, Jason Droege, blaming the adjustments partially on “shifts in marketplace call for.” Droege stated Scale AI would workforce up in different portions of the industry, together with executive gross sales — the corporate simply landed a $99 million contract with the U.S. Military.
Some speculated to begin with that Meta’s funding in Scale AI was once truly to entice Wang, a founder who has operated within the AI house since Scale AI was once based in 2016 and who seems to be serving to Meta to draw most sensible AI skill.
Except Wang, there’s an open query round how treasured Scale is to Meta.
One present MSL worker says that a number of of the Scale executives introduced over to Meta don’t seem to be running at the core TBD Labs workforce, as with Mayer. Additional, Meta isn’t solely depending on Scale AI for information labeling paintings.
In the meantime, Meta’s AI unit has grow to be more and more chaotic since bringing on Wang and a wave of most sensible researchers, consistent with two former workers and one present MSL worker. New skill from OpenAI and Scale AI have expressed frustration with navigating the paperwork of a giant corporate, whilst Meta’s earlier GenAI workforce has observed its scope restricted, they stated.
The tensions point out that Meta’s biggest AI funding up to now could also be off to a rocky get started, regardless of that it was once intended to deal with the corporate’s AI construction demanding situations. After the lackluster release of Llama 4 in April, Meta CEO Mark Zuckerberg grew annoyed with the corporate’s AI workforce, one present and one former worker instructed TechCrunch.
So that you could flip issues round and meet up with OpenAI and Google, Zuckerberg rushed to strike offers and introduced an competitive marketing campaign to recruit most sensible AI skill.
Past Wang, Zuckerberg has controlled to drag in most sensible AI researchers from OpenAI, Google DeepMind, and Anthropic. Meta has additionally got AI voice startups together with Play AI and WaveForms AI, and introduced a partnership with the AI symbol technology startup, Midjourney.
To energy its AI ambitions, Meta not too long ago introduced a number of large information middle buildouts around the U.S. Some of the biggest is a $50 billion information middle in Louisiana referred to as Hyperion, named after a titan in Greek mythology that fathered the God of Solar.
Wang, who’s now not an AI researcher through background, was once considered as a rather unconventional selection to steer an AI lab. Zuckerberg reportedly held talks to herald extra conventional applicants to steer the hassle, corresponding to OpenAI’s leader analysis officer, Mark Chen, and attempted to procure the startups of Ilya Sutskever and Mira Murati. They all declined.
Probably the most new AI researchers not too long ago introduced in from OpenAI have already left Meta, Stressed in the past reported. In the meantime, many longtime participants of Meta’s GenAI unit have departed in mild of the adjustments.
MSL AI researcher Rishabh Agarwal is without doubt one of the newest, posting on X this week that he’d be leaving the corporate.
“The pitch from Mark and @alexandr_wang to construct within the Superintelligence workforce was once extremely compelling,” stated Agarwal. “However I in the long run select to observe Mark’s personal recommendation: ‘In an international that’s converting so speedy, the most important chance you’ll be able to take isn’t taking any chance’.”
Requested in a while about his time at Meta and what drove his determination to depart, Agarwal declined to remark.
Director of product control for generative AI, Chaya Nayak, and analysis engineer, Rohan Varma, have additionally introduced their departure from Meta in fresh weeks. The query now’s whether or not Meta can stabilize its AI operations and retain the skill it wishes for its long run good fortune.
MSL has already got to work on its subsequent technology AI style. In keeping with studies from Industry Insider, it’s aiming to release it through the tip of this yr.
