• Meta formed a new AI unit, Meta Superintelligence Labs.
• It follows Llama 4 backlash and staff exits as Meta pivots to AGI.

Meta has regrouped its artificial intelligence operations under a new unit called Meta Superintelligence Labs (MSL), according to a source familiar with the matter. The decision comes after the company’s latest open-source model, Llama 4, received poor feedback, and several senior researchers left. With competitors like OpenAI, Google, and China’s DeepSeek gaining ground, Meta is looking to reset its AI strategy and rebuild momentum.

The new group will be led by Alexandr Wang, the former CEO of Scale AI, who joins Meta as its new chief AI officer. Wang made his name building Scale into a key data provider for many large AI models. He won’t be working alone—Nat Friedman, ex-CEO of GitHub and a known investor in AI startups, will co-lead MSL and focus on product development and applied research. According to the source, Friedman has been advising Meta over the past year and is already familiar with its roadmap.

Meta CEO Mark Zuckerberg is personally backing the shake-up. Over the past month, he’s taken a direct role in recruiting, sending messages to candidates via WhatsApp and courting several startup founders. One of those was Safe Superintelligence (SSI), co-founded by OpenAI’s Ilya Sutskever. Although a deal with SSI didn’t go through, its co-founder and CEO, Daniel Gross, is said to have joined MSL.

In early June, Meta also invested $14.3 billion into Scale AI, further tying the company to its new leadership. Zuckerberg hopes the reshaped team will accelerate Meta’s push toward artificial general intelligence (AGI)—a still-hypothetical form of AI that can outperform humans in most tasks—and help create new revenue streams from tools like Meta AI, image-to-video ad generators, and smart wearables.

MSL is already pulling in talent from top AI labs. The source said Meta has made at least 11 new hires in recent weeks, including names from OpenAI, Google DeepMind, Anthropic, and DeepSeek. Some of the individuals joining include:

• Jack Rae and Pei Sun, both previously with DeepMind
• Joel Pobar, returning to Meta after time at Anthropic
• Jiahui Yu, Shuchao Bi, Shengjia Zhao, and Hongyu Ren from OpenAI
• Huiwen Chang, who worked on image generation at Google Research
• Trapit Bansal, known for work on reasoning in AI models
• Ji Lin and Johan Schalkwyk, both with backgrounds in model architecture and voice systems

These hires come after OpenAI CEO Sam Altman said Meta had offered some of his staff bonuses of up to $100 million to switch companies.

Despite the influx of talent, some analysts are sceptical about how soon Meta will see results. Its last big technology bet, Reality Labs, has cost the company more than $60 billion since 2020, with only a few products—such as Ray-Ban smart glasses and Quest headsets—making it into users’ hands.

AGI remains a long-term goal for the tech industry. Microsoft recently spent $650 million to hire most of Inflection AI’s staff, including co-founder Mustafa Suleyman, while Amazon has been hiring AI researchers from startups like Adept. But the path to AGI is unclear. Meta’s own chief AI scientist, Yann LeCun, has said the current tools and approaches won’t be enough. Meanwhile, SoftBank’s Masayoshi Son believes a breakthrough could happen within a decade.

In an internal memo on Monday, Zuckerberg said Meta is now focused on building “personal superintelligence for everyone.” The memo confirmed that MSL will combine Meta’s foundation model research, FAIR team, and AI product development under one roof. A new lab within MSL will also start work on a next-generation model aimed at pushing the current technical limits.

Zuckerberg wrote that Meta is “uniquely positioned” to take this step due to its large user base, experience running global apps, and capacity to invest in large-scale computing. He also pointed to Meta’s early work on AI-powered wearables as another potential edge.

The memo ended with a call for more staff to join the new lab, and suggested more hires will be announced in the coming weeks.

Whether this new approach helps Meta catch up with its competitors—or delivers anything close to AGI—remains to be seen.