Zuckerberg signals Meta won’t open source all of its ‘superintelligence’ AI models

In a recent address, Meta CEO Mark Zuckerberg unveiled his vision for what he terms 'personal superintelligence'—the concept that individuals should harness AI to fulfill their personal aspirations. This announcement subtly indicates a pivotal change in Meta's approach to releasing AI models as the company ambitiously ventures into the realm of superintelligence. Zuckerberg emphasized the importance of sharing the advantages of superintelligence broadly, while also highlighting the unique safety challenges it presents. He stated, "We’ll need to be rigorous about mitigating these risks and careful about what we choose to open source." This remark carries weight, especially considering Zuckerberg's previous commitment to the Llama series of open models, which has differentiated Meta from rivals such as OpenAI, xAI, and Google DeepMind. The overarching goal has been to develop open AI models that rival or surpass their closed counterparts. In a letter penned in 2024, Zuckerberg expressed optimism about the future of the Llama models, predicting they would soon be industry leaders. However, critics argue that Llama does not fully adhere to the principles of open source AI, particularly since Meta has not disclosed its vast training datasets. This evolving narrative suggests that Meta may be reconsidering its default stance on open source as it navigates the competitive landscape. Zuckerberg's remarks also shed light on the motivations behind closed models, which grant companies greater control over their monetization strategies. He previously pointed out that Meta's business model does not depend on selling access to AI models, allowing the company a unique position. As Meta strategizes to catch up with competitors, the pressure is mounting, especially with the development of the Llama 3 model. The urgency intensified in June 2025, when Meta committed a staggering $14.3 billion to Scale AI, signaling its serious intent to accelerate its artificial general intelligence (AGI) initiatives. This investment included acquiring key talent and establishing a new division, Meta Superintelligence Labs, dedicated to AI advancements. Recent developments suggest that Meta may have put a hold on the testing of its latest Llama model, dubbed Behemoth, and is now redirecting its focus towards creating a closed model. Zuckerberg's ambition to integrate 'personal superintelligence' into consumer products like augmented reality glasses and virtual reality headsets is taking shape. He articulated that devices such as glasses equipped with context-awareness capabilities will become central to our computing experiences. In response to inquiries regarding the potential for keeping advanced models proprietary, a Meta spokesperson reaffirmed the company's commitment to open source AI while also indicating a future that includes a blend of open and closed models. "We plan to continue releasing leading open source models," the spokesperson noted, emphasizing a balanced approach moving forward.