Meta Platforms CEO Mark Zuckerberg is advocating for reduced barriers in the United States regarding open-source AI to enhance competition against Chinese competitors. The social media giant unveiled a smaller, more customizable AI model named Muse Glimmer, designed to operate on a single graphics card on Mac or PC, catering to the demand for AI systems running directly on users’ devices.
Zuckerberg expressed plans for even larger models in the future, emphasizing the importance of democratizing AI access rather than centralizing power. The endorsement of open-weight AI by Zuckerberg aligns with a growing trend among businesses concerned about escalating AI costs and recent cybersecurity issues related to models like Kimi K3 and incidents involving companies such as Anthropic, OpenAI, and Meta.
The preference for open-weight models stems from their customizable nature, allowing users to tailor the models for specific needs, unlike closed models controlled entirely by companies. Notably, Chinese startups have taken the lead in open-weight models, with offerings like Kimi K3, Qwen3.8-Max, and V4-Flash competing against established U.S. counterparts.
Meta’s stock, which has experienced a 10% decline this year, saw a nearly 3% increase in premarket trading following the AI model developments. The company is set to release Muse Spark 1.2, its advanced model, developed by a specialized team formed to boost its AI capabilities after setbacks with previous models.
In a comprehensive essay on AI, Zuckerberg announced a $1 billion fund aimed at addressing concerns surrounding data center expansions and supporting local communities affected by Meta’s infrastructure development. Amidst challenges in infrastructure development in the U.S., Zuckerberg stressed the need for policy revisions to empower domestic firms in the open-weight AI arena, particularly in data use and distillation techniques.
Zuckerberg’s proposal includes implementing a governance framework for independent directors to approve model releases based on safety criteria. The U.S. administration’s stance on not subjecting open-weight AI models to voluntary safety tests was also highlighted, emphasizing the need for streamlined regulations to support the advancement of American open-source AI models.
