Meta Platforms CEO Mark Zuckerberg has urged the United States to reduce barriers on open-source AI to enhance competition with Chinese counterparts. The company unveiled a new AI model named Muse Glimmer, which is smaller and more customizable than rival models. Zuckerberg hinted at the release of larger models in the future to cater to the increasing demand for AI systems run directly on personal devices.
The push for open-weight AI reflects the industry’s shift towards customizable and cost-effective models, as concerns grow over the security of closed models. Recent incidents involving Chinese open-weight models and cybersecurity breaches have highlighted the advantages of open-source platforms.
Meta, along with other tech giants like Nvidia and Microsoft, has pledged support for open-weight AI. Chinese startups have been leading in this field, showcasing competitive models like Moonshot’s Kimi K3 and Alibaba’s Qwen3.8-Max. In contrast, U.S. developers have predominantly focused on closed weight models.
In a bid to ease community concerns, Meta plans to release its most advanced model, Muse Spark 1.2, and invest in community support initiatives. Zuckerberg emphasized the need for policy reforms to ensure American leadership in open-weight models, including data use regulations and model distillation techniques.
The company’s commitment to transparency and community engagement comes amidst a broader debate on data center expansion and infrastructure development. Zuckerberg highlighted the challenges the U.S. faces in building AI infrastructure compared to other countries, calling for a reevaluation of current policies to foster innovation in the AI sector.
Despite the current administration’s stance on safety testing for open-weight AI models, industry leaders continue to advocate for a more open and collaborative approach to AI development. Zuckerberg’s call for reducing barriers and embracing open-source AI reflects a broader industry trend towards democratizing AI technologies.