With new open models, Meta pitches another reboot of its struggling AI strategy

Meta CEO Mark Zuckerberg outlined a vision for the United States to compete against China in the global race to dominate artificial intelligence (AI) in a recent essay. He emphasized the importance of preventing government overreach and ensuring that superintelligence technology becomes accessible to everyone. Superintelligence refers to the point at which AI surpasses human intelligence.

To achieve this, Zuckerberg advocated for open weight models, which allow users to download the rules governing an AI system's behavior. Unlike open-source software that makes all underlying code publicly available, open weight models do not disclose their entire codebase. Meta has invested billions of dollars into its AI division, Meta Superintelligence Labs, to stay competitive with rivals like OpenAI and Anthropic, which develop closed AI models.

In response to the growing competition from Chinese companies such as Alibaba, DeepSeek, and Moonshot, Meta has unveiled a new open weight model called Glimmer. This model, derived partly from a closed-source model called Muse Spark, requires less computing power than other models but remains powerful enough to operate AI agents or software capable of performing tasks without human intervention.

In his essay, Zuckerberg also criticized companies focused solely on building AI for specific institutions, arguing that this approach could concentrate power in the hands of entities rather than individuals. He stressed the need to distribute superintelligence widely and empower every individual to direct it. With a user base of 3.6 billion across its apps, including Facebook, Instagram, and WhatsApp, Meta faces increasing concerns over app usage by American youths, leading to a recent $1 billion damage award in a New Mexico case.

Additionally, jury selection is set to begin in a California court for a trial brought by four U.S. states seeking $1.4 trillion in damages from Meta. The company is also navigating a growing debate about AI regulations in the U.S. and local opposition to data center development. In June, President Donald Trump ordered the establishment of a voluntary review process for advanced AI models before their release, but the administration missed its deadline.

Zuckerberg believes that a flexible regulatory environment for open models would benefit Meta, OpenAI, and Anthropic, as well as help the U.S. compete with Chinese developers. To ensure American open source models remain globally competitive, Zuckerberg emphasized the importance of removing obstacles that hinder their competitiveness.