TLDR: Mark Zuckerberg, CEO of Meta, has indicated a potential shift in the company’s open-source strategy for its most advanced “superintelligence” AI models, citing novel safety concerns. This marks a departure from Meta’s previous commitment to open-sourcing AI, including its Llama models, as the company aims to rigorously mitigate risks associated with increasingly capable AI.
Meta CEO Mark Zuckerberg has signaled a potential change in the company’s approach to open-sourcing its most advanced “superintelligence” AI models. In a letter published on Wednesday, July 30, 2025, Zuckerberg outlined his vision for “personal superintelligence,” where AI empowers individuals to achieve personal goals and enhance their lives. However, he also indicated that Meta’s most sophisticated AI might remain closed to ensure the company retains control and can effectively address emerging safety concerns.
This stance represents a notable shift for Meta, which has historically championed open AI models like Llama, differentiating itself from competitors such as OpenAI, xAI, and Google DeepMind. In a 2024 letter, Zuckerberg had expressed ambitions for future Llama models to become industry leaders. Yet, he had previously hinted at limitations, stating in a past podcast, “If at some point however there’s some qualitative change in what the thing is capable of, and we feel like it’s not responsible to open source it, then we won’t.”
Zuckerberg emphasized that while “the benefits of superintelligence should be shared with the world as broadly as possible,” these advanced systems will “raise novel safety concerns.” He stressed the need for Meta to be “rigorous about mitigating these risks and careful about what we choose to open source.”
Reports suggest that Meta may be re-evaluating its strategy amidst the competitive landscape, with some sources indicating a halt in testing for its latest Llama model, Behemoth, to prioritize the development of closed models. This strategic pivot follows significant investments, including $14.3 billion in Scale AI, the acquisition of its founder, and the establishment of a specialized unit called Meta Superintelligence Labs.
Also Read:
- Meta’s Strategic $40 Billion AI Investment and Open-Source Commitment Face Wall Street Scrutiny
- Meta’s Billion-Dollar AI Talent Hunt Hits Roadblock at Thinking Machines Lab
Despite this potential shift, a Meta spokesperson confirmed the company’s continued support for open-source AI development. Zuckerberg’s vision for “personal superintelligence” is ultimately aimed at consumers through products like augmented reality glasses and virtual reality headsets, which he believes will become primary computing devices by understanding user context and interacting throughout the day.
