MENLO PARK, California — Meta CEO Mark Zuckerberg has outlined an ambitious vision for artificial intelligence, calling for the United States to maintain its lead in a global race with China while arguing that advanced AI should be widely accessible rather than controlled by governments or large institutions.
In a lengthy essay published Monday, Zuckerberg said the development of so-called superintelligence — a theoretical level of AI capability that surpasses human intelligence — could reshape the balance of power globally.
He argued that the United States should promote open-weight AI models, which allow users to download and run models while accessing the weights that determine their behaviour.
Unlike fully open-source software, however, open-weight models do not necessarily make all underlying source code publicly available.
Meta Bets on Open AI Models
Zuckerberg’s position places Meta in contrast with companies such as OpenAI and Anthropic, whose leading AI models are largely developed and operated as closed systems.
Meta has invested billions of dollars in artificial intelligence through its Meta Superintelligence Labs as it seeks to narrow the gap with OpenAI, the developer of ChatGPT, and Anthropic, the company behind Claude.
Meta also announced a new open-weight model called Glimmer, which it said was developed partly using technology from Muse Spark, a closed-source model announced earlier this year.
According to Meta, Glimmer requires less computing power than some competing systems while remaining capable of operating AI agents that can perform tasks with limited human intervention.
The company sees open models as a way to encourage wider adoption among developers while strengthening the position of American AI companies against increasingly competitive Chinese developers.
Growing Competition From China
Zuckerberg’s argument comes as Chinese technology companies, including Alibaba, DeepSeek and Moonshot, have developed increasingly capable and relatively inexpensive open AI models.
The growing competition has pressured US companies to develop cheaper systems and reduce the cost of accessing advanced AI capabilities.
Zuckerberg also criticised an approach in which advanced AI is primarily developed for corporations, governments and other institutions.
“Rather than centralizing superintelligence, we should distribute it widely and give every person the ability to direct it,” Zuckerberg wrote.
He argued that concentrating advanced AI in the hands of institutions could shift power away from individuals.
OpenAI and Anthropic have both secured government contracts, adding another dimension to the debate over who should control increasingly powerful AI systems.
Regulatory Pressure Builds
Meta’s push comes as governments grapple with how to regulate rapidly advancing AI technology.
US President Donald Trump ordered the creation of a voluntary review process for advanced AI models in June, although his administration missed an August 1 deadline to provide details of the proposed framework.
Zuckerberg called for a flexible regulatory approach that would allow US AI companies to innovate while competing internationally.
“Our goal should be for American open source models to be the best globally,” he wrote.
He warned that regulatory delays could undermine American competitiveness by giving foreign AI developers an opportunity to move faster.
At the same time, Meta is facing growing scrutiny over the social impact of its platforms. The company has recently faced major legal challenges related to alleged harms to children using Facebook and Instagram, including a New Mexico case involving nearly $1 billion in damages.
Meta also faces a separate multistate lawsuit in California seeking potentially $1.4 trillion in damages.
Meta Expands AI Investment
Meta operates Facebook, Instagram and WhatsApp, with about 3.6 billion users across its platforms.
The company is nevertheless shifting substantial resources toward AI infrastructure and research. Meta expects to spend as much as $145 billion this year, nearly twice its 2025 expenditure, as it accelerates investment in AI.
The company is also responding to opposition to the construction of data centres needed to power AI systems. A Meta spokesperson said the company plans to establish a $1 billion fund to support communities around its facilities.
