Meta has been making headlines for trying to poach researchers from rival AI labs with eye-popping offers as it races to build “personal superintelligence.”
The effort has supercharged Silicon Valley’s talent wars, and it’s also rankling some of the AI researchers who joined Meta before its superintelligence push, according to current and former Meta employees who requested anonymity for fear of retaliation.
The grumblings are particularly acute at the generative AI team within Meta, known as GenAI. This group worked on Llama 4, an AI model that got a lukewarm reception when it was released earlier this year.
“Meta has really good researchers. Their treatment of researchers who are not part of superintelligence has been sub-par,” reads a now-deleted X post from Rohan Anil, who worked on Llama 4 for Meta. “It feels like a giant social experiment.”
Anil left Meta and joined the AI startup Anthropic in June, shortly before Meta CEO Mark Zuckerberg publicly launched this new, more intense phase of the AI talent wars. Anil declined to comment.
“People join Meta because they want to work with the best in the industry, have access to industry-leading levels of compute, and develop superintelligence,” a Meta spokesperson told Business Insider. “Those pushing certain narratives typically have ulterior motives.”
Meta has the best retention rates in the Magnificent 7 and is growing engineering teams two to three times faster than it’s losing them, according to a report from the VC firm SignalFire. Meta has poached well over a dozen leading AI researchers, mostly from OpenAI, such as the ChatGPT cocreator Shengjia Zhao, but also from other rivals.
Feeling like a failure
Among some of Meta’s existing AI staff, the new superintelligence team is causing rifts, according to the current and former employees who spoke with BI.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Bringing in outsiders for 10 to 50 times higher compensation made it feel like Zuckerberg was telling GenAI employees they had failed, a former Meta AI researcher said.
Some AI employees are considering leaving to join rival labs, while others are threatening to quit so they can negotiate a spot on Meta Superintelligence Labs, or MSL, which is causing some chaos at the company, this researcher added.
Meta told BI it doesn’t counter offers for employees who threaten to leave, and it hasn’t done so in the past.
FAIR or unFAIR?
The tensions aren’t just about compensation but also revolve around access to computing resources and the prestige of being associated with the elite team at the center of MSL, the researcher added.
Though MSL now oversees the entire GenAI team — which itself was reorganized in May — MSL maintains a core unit of star researchers dubbed the “TBD Lab,” The Wall Street Journal reported.
Other AI labs are seeing opportunity from this internal Meta tension. Elon Musk’s xAI has hired more than a dozen Meta researchers, with some joining as recently as the past few weeks, while Microsoft is also trying to poach Meta AI talent, BI previously reported.
FAIR, Meta’s long-standing AI research lab that’s also now under MSL, is less directly affected by the tensions, according to an employee there, and has largely retained its independence.
Yann LeCun, FAIR’s chief scientist, is focused on advising Zuckerberg, speaking at conferences, and leading his own small team of about 10 researchers focused on building what he believes is the key to artificial general intelligence, a model called I-JEPA, this employee told Business Insider.
A departure and an experiment
On August 6, Laurens van der Maaten, a distinguished research scientist at Meta, announced he was joining Anthropic. His LinkedIn shows he left Meta in June before the formal creation of MSL. Van der Maaten didn’t respond to a request for comment.
Reacting to the move on X, Erik Meijer, a senior engineering director at Meta until 2024, wrote: “Every action has a reaction; the unintended side effects of creating a SI team,” referring to Meta’s superintelligence team.
When BI contacted Meijer for comment, he responded by sending a YouTube clip of an experiment in which two monkeys perform the same task but receive different rewards. The one given the less tasty treat hurls it back at its handler and angrily shakes its cage.
Meta AI insiders who spoke with BI weren’t necessarily skeptical of Zuckerberg’s superintelligence push and said the recruiting campaign made sense from the CEO’s perspective.
One said the GenAI team simply grew too fast after the launch of ChatGPT, making it difficult to manage. Another suggested that if Zuckerberg wants to start anew, it may be a good thing if people leave the company.
Have a tip? Contact this reporter via email at crollet@insider.com or on Signal and WhatsApp at 628-282-2811. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.
