Meta CEO: US Must Lower AI Barriers for Open-Source

Meta CEO Mark Zuckerberg has urged the United States to reduce regulatory hurdles for open-source artificial intelligence models, a move he believes is crucial for domestic firms to compete effectively with international rivals, particularly from China.

Zuckerberg's call coincides with Meta's release of Muse Glimmer, a new, smaller open-weight AI model designed for agentic tasks and capable of running on consumer-grade hardware like a Mac or PC. This development aims to capitalize on the growing demand for AI systems that operate directly on user devices.

The social media giant is bolstering its AI capabilities through its dedicated superintelligence team, formed last year. Zuckerberg highlighted that American labs face more restrictions on training data compared to foreign competitors, advocating for policy adjustments to support U.S. leadership in open-weight AI.

He also emphasized the importance of AI model distillation, where powerful AI systems train smaller ones, and proposed that Meta will implement a governance structure empowering independent directors to approve safety criteria for model releases.