Meta’s latest artificial intelligence hiring spree, backed by massive compensation packages, is fueling internal friction between new recruits and longtime researchers as the company reorganizes its AI efforts, according to a report.
Costly hires and a strategic pivot
According to the New York Post, which cites reporting by the New York Times, CEO Mark Zuckerberg’s recent moves have led to clashes within the AI division. Alexandr Wang — the 28-year-old billionaire who led Scale AI before Meta acquihired him as chief AI officer — is reportedly in a standoff with veteran engineers over the company’s AI direction. The report says Wang’s superintelligence team is pushing to make Meta’s next model “closed,” a departure from Meta’s tradition of open-sourcing its models.
The new leadership has abandoned a previous frontier effort called Behemoth after disappointing tests delayed its release last spring, the report said. Meta invested $14.3 billion in Scale AI before bringing Wang aboard to lead the company’s push against rivals.
Shengjia Zhao, an OpenAI veteran who co-created ChatGPT and was named Meta’s chief AI scientist in July, has been rigorously assessing longtime employees for roles in the restructured organization. Researchers have reportedly lined up outside Zhao’s office in recent weeks as he conducts evaluations.
Reorg, departures, and spending plans
The AI division is undergoing a reorganization into four groups focused on research, superintelligence development, products, and infrastructure, according to the report. Meta is also considering downsizing the AI unit, which has grown to thousands of employees, through potential role eliminations and transfers to other departments, the Post reported.
Several departures have followed: Joelle Pineau, a leading computer scientist at Meta, left and later joined Cohere; Angela Fan, who worked on the open-source Llama model, moved to OpenAI; and Loredana Crisan, vice president of generative AI, is leaving for Figma as chief design officer, according to the Times as cited by the Post. Former GitHub CEO Nat Friedman and Daniel Gross, who previously ran Safe Superintelligence, will lead development of new AI features under products and applied research.
Open vs. closed models and external tech
The report says Meta is exploring the use of third-party AI models to power products, either by building on other open-source models or licensing closed-source technology — a shift from its prior practice of relying on its own systems. Meta spokesman Andy Stone mocked media coverage of the shakeup, accusing reporters of exaggerating internal restructuring. The company declined to comment on the reorganization, which was first partially reported by The Information.
Zuckerberg recently pitched investors on “a new era of individual empowerment” while projecting capital expenditures up to $72 billion this year, as the company races to compete in AI.
