Meta pushes global AI vision amid race with OpenAI, Anthropic and China

San Francisco: Meta CEO Mark Zuckerberg outlined a global AI strategy in a Monday essay, emphasizing the need for the United States to compete against China and prevent government control over superintelligence, a point where AI surpasses human intelligence. He advocated for open-weight models, which allow users to download governing rules for AI systems, despite not revealing their underlying code.

In contrast, competitors like OpenAI and Anthropic focus on closed models, often referred to as "black boxes." Meta has invested billions in its AI division, Meta Superintelligence Labs, to stay competitive with OpenAI and Anthropic, the developers of ChatGPT and Claude, respectively. The tech giant also unveiled Glimmer, an open-weight model derived from a closed-source model and updated last week, designed to operate AI agents efficiently.

Chinese firms, including Alibaba, DeepSeek, and Moonshot, have developed advanced open models, prompting American companies to reduce prices and create cheaper alternatives. Zuckerberg criticized labs that focus on AI for institutions, arguing it tips power to entities rather than individuals. He proposed distributing superintelligence widely to give every person the ability to direct it.

Meta's 3.6 billion users, comprising Facebook, Instagram, and WhatsApp, have raised concerns about app usage among American youths. The company faces a $1 billion lawsuit in New Mexico and a potential $1.4 trillion claim from four U.S. states in California. Meanwhile, the U.S. government is debating AI regulation, with President Trump setting up a voluntary review process for advanced AI models.

Zuckerberg suggested a collaborative, flexible regulatory environment would benefit American open-source models, enabling them to compete globally and potentially outrun foreign developers.