It is only since June that Meta invested $ 14.3 billion in data-labeling seller Scale AI, CEO Alexandra Wang and many top officials of the startup brought Meta Superintendent Labs (MSL) to run. But the relationship between the two companies is already indicating horrors.
Wang, one of the least officers, brought to help MSL run – Former Senior Vice President of Scale AI, Jenai, former Vice President of Product and Operations, Ruben Mayor, along with the company, sent Meta with the company only two months later, two people familiar with the matter told Tekkachch.
The mayor spent almost five years with Scale AI in two stints. In its short time in the meta, according to the sources, the mayor looked at the AI data operating teams, but was not part of the company’s TBD labs – the core unit inside the meta worked with the construction of the AI superintendent, where the top AI researchers of the OpenaiI have landed.
However, the Mayor disputed some details about his role, told Techcrunch that his initial position was “to help install the lab, whatever was necessary” instead of data, and he was “part of TBD Labs” from Day One instead of being “excluded from the core AI unit. [Wang]”And was” very happy “with his meta experience.
Beyond the change of personnel, Meta’s relationship with Scale AI seems to be shifting. TBD Labs is working with third -party data labeling vendors in addition to Scale AI to train its upcoming AI model, according to five people familiar with the case. Those third-party vendors include Mercer and Serge, two of the two biggest contestants of Scale AI, people said.
While AI Labs usually work with many data labeling vendors – Meta is working with Merkor and Serge because before finishing TBD labs – it is rare to invest so heavy in a data seller for AI Lab. This makes this situation particularly notable: even with the investment of Multi-Billion-Dollar of Meta, several sources said that researchers in TBD labs have seen the data of Scale AI as a low quality and expressed priority to work with Surge and Mercer.
Scale AI initially created its business on a crowdsourcing model, with a large, low -cost workforce to handle simple data labeling, which is the process of tagging and anotating raw information to train AI models. But as the AI models have become more sophisticated, they now require high-skilled domain experts-like doctors, lawyers and scientists-to generate and refine the high quality data required to improve their performance.
Techcrunch event
San francisco
,
27-29 October, 2025
Although Scale AI has moved to attract these subject matter experts with its external platform, competitors such as Serge and Merkor are growing quickly as their business models were built on the foundation of high-paying talent from the beginning.
A Meta spokesperson disputed the fact that Scale AI had quality issues with products. Serge and Mercer refused to comment. When asked about the deeper dependence of the meta on competitive data providers, a scale AI spokesperson directed Techchan for his initial announcement of investing Meta in startups, creating expansion of commercial relations of companies.
Meta’s third-party data with data vendors means that the company is not laying all its eggs in Scale AI even after investing billions in startups. However, this cannot be said for Scale AI. After Meta announced his massive investment with Scale AI, Openi and Google said they would stop working with the data provider.
Shortly after losing those customers, Scale AI put 200 employees in its data labeling business in July, with the company’s new CEO, Jason Draws, blamed the change in the part on the “change in market demand”. Draws stated that Scale AI would do staff in other parts of the business, including government sales – the company signed a $ 99 million with the US Army.
Some people initially speculate that Meta’s investment in Scale AI was actually to woo Wang, a founder who operated in AI Space as Scale AI was established in 2016 and helps Meta to attract the top AI talent.
Apart from Wang, there is an open question of how valuable the valuable scale for meta is.
A current MSL employee says that several scale officers brought to Meta are not working in the Core TBD Labs team.
Meanwhile, according to two former employees and an existing MSL employee, the AI unit of the meta has become increasingly chaotic since bringing a wave of Wang and top researchers. He said that the new talent of Openai and Scale AI has expressed disappointment in navigating the bureaucracy of a large company, while the previous Meta’s previous Genai team has seen their scope limited.
Stress indicates that the largest AI investment of meta so far may stop for a rocky start, yet it was the company’s AI development challenges. After the shortage of Lama 4 in April, Mata CEO Mark Zuckerberg was disappointed with the company’s AI team, a current and a former employee told Techchchan.
In an attempt to turn and catch things with Openai and Google, Zuckerberg raced to attack deals and launched an aggressive campaign to recruit top AI talent.
Beyond Wang, Zuckerberg has succeeded in pulling the top AI researchers of Openi, Google Dipmind and Anthropic. Meta has also acquired AI Voice Startups including Play AI and Waverform AI, and announced a partnership with AI image generation startups, midzorney.
To strengthen its AI ambitions, Meta recently announced several large -scale data center buildouts in the US, one of the largest Louisiana, Louisiana has a $ 50 billion data center, called Hyperian, named Hyperian, named after a Titan in Greek mythology, which gives birth to the Sun God.
Wang, who is not AI researchers by the background, was seen to be somewhat unconventional options to lead the AI lab. Zuckerberg allegedly talked about bringing more traditional candidates, such as OpenaiI’s Chief Research Officer, Mark Chen, and Ilya Sutaskewar and Meera Murati’s startups. All of them refused.
Some new AI researchers recently brought from Openai have already left the meta, Wired had earlier told. Meanwhile, many long members of the Meta Jenai unit of Meta have gone to light of changes.
MSL AI researcher Rishabh Aggarwal is posting on X this week, posting on X this week that he will leave the company.
Aggarwal said, “The pitch from Mark and @Alexandr_wang was incredibly compelling to build in the superintendent team.” “But I finally choose to follow Mark’s advice: ‘In a world that is changing so fast, the biggest risk you can take no risk.”
Later, Meta was asked about his time and left his decision to leave, Aggarwal refused to comment.
The director of Product Management for Generic AI, Chaya Nayak, and Research Engineer, Rohan Verma has also announced his departure from Meta in recent weeks. Now the question is whether Meta can stabilize its AI operations and maintain the talent required for its future success.
MSL has already started working on its next generation AI model. According to Business Insider reports, it aims to launch it by the end of this year.
Update: This story has been updated with Mayor’s comments, which reached Techcrunch after publication.
