Meta Platforms, Inc. CEO Mark Zuckerberg has been making adjustments to the company's artificial intelligence operations over recent months. He now plans to implement further changes that could intensify internal upheaval within this technology sector.

According to sources familiar with the matter, Meta Platforms, Inc. is expected to announce on Tuesday that it will divide its AI division—the Meta Superintelligence Lab—into four separate groups. One group will concentrate on artificial intelligence research; another will focus on potentially powerful AI technology called "superintelligence"; a third group will handle product-related work; and the fourth will focus on infrastructure such as data centers and other AI hardware. These initiatives aim to better integrate Meta's resources and enable faster AI product development to compete with other companies, the sources indicated.

Sources suggest that some AI division executives are anticipated to depart. Additionally, given that the department has expanded to thousands of employees in recent years, Meta Platforms, Inc. is considering overall streamlining of the AI division, which may include layoffs or transferring staff to other company departments. They also mentioned that related discussions are ongoing and no final decisions have been made.

According to sources, Meta Platforms, Inc. previously relied solely on proprietary technology to power its AI products, but the company is now actively exploring the use of third-party AI models, including development based on other freely available "open-source" AI models or licensing "closed-source" models from other companies.

These changes follow months of turbulence and reorganization surrounding Meta's AI strategy. The 41-year-old Zuckerberg has been willing to invest heavily and undertake comprehensive company reforms to maintain competitiveness in the AI field—where competition to build the most advanced AI technology has narrowed to a few key players. As the AI race continues to produce new winners and losers, Meta's performance is under close scrutiny.

In June of this year, Zuckerberg's determination became increasingly evident after Meta Platforms, Inc. struggled to advance development of its latest AI model. That month, Meta announced the establishment of the Superintelligence Lab, specifically dedicated to creating AI more powerful than the human brain. Meta invested $14.3 billion in startup Scale AI and hired the company's CEO Alexander Wang as its new Chief AI Officer. Additionally, Meta offered compensation packages worth hundreds of millions of dollars to recruit researchers from competitors like OpenAI and Google, triggering a "talent war" in Silicon Valley.

Alexander Wang, Meta's Chief AI Officer, whose new team has created tensions within the company.

During last month's investor conference call, Zuckerberg stated he was betting that superintelligence would usher in a "new era of individual empowerment," adding that AI has already improved Meta's core advertising business. The company indicated that capital expenditure this year could reach as high as $72 billion, with most allocated to data center construction and AI researcher recruitment.

A Meta spokesperson declined to comment. Some details of this restructuring were previously reported by The Information.

Tensions have been apparent since Zuckerberg established the superintelligence team under Alexander Wang's leadership. Sources revealed that Wang's team is focusing on building the company's most powerful AI model—the so-called "frontier model."

The new team has discussed making Meta's next-generation AI model "closed-source," which would represent a significant departure from the company's long-held "open-source" model philosophy. Closed-source models strictly protect their underlying code, while open-source AI models allow other developers to build upon them.

Sources indicated that the new team has decided to abandon Meta's previous frontier model called "Behemoth" and start developing a new model from scratch. One person stated that last spring, the Behemoth model's release was delayed due to poor performance test results.

As Meta Platforms, Inc. spends billions of dollars bringing in AI talent, three sources revealed that some company veterans are dissatisfied with these new hires.

In July, Meta appointed OpenAI researcher and ChatGPT chatbot co-creator Shengjia Zhao as Chief AI Scientist. According to one source, in recent weeks, a group of Meta's veteran AI researchers and employees have been queuing outside Zhao's office as he interviews them for new positions, asking about their past work.

Sources indicated that former GitHub CEO Nat Friedman and Daniel Gross, who previously operated a startup called "Safe Superintelligence," will lead the development of new AI features in the product and application research division.

Personnel changes continue to occur. Earlier this year, Meta's top computer scientist Joelle Pineau departed and subsequently joined AI startup Cohere. Research scientist Angela Fan, who participated in developing Meta's open-source AI model Llama, recently also moved to OpenAI. Sources indicated that Vice President of Generative AI Loredana Crisan is expected to announce her departure on Tuesday.

However, some senior AI leaders remain with the company. Sources indicated that Rob Fergus, who helped establish Meta's AI research division in 2014, will continue as the day-to-day leader of Meta's Fundamental AI Research lab (FAIR), which promotes AI technology development through open-source research. Ahmad Al-Dahle and Amir Frenkle, who previously worked on generative AI product development, now report to Alexander Wang, focusing on strategic AI projects.