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

Meta boss Mark Zuckerberg published a lengthy essay on Monday outlining a new Meta AI vision, urging the United States to outcompete China, resist government control over artificial intelligence and open superintelligence to everyone. Meta also unveiled Glimmer, an open weight model built to power AI agents at lower computing cost.

What is Meta doing in the AI race?

Meta is pouring billions into Meta Superintelligence Labs and promoting open weight models, which let users download the rules governing an AI system's behavior. Unlike the closed models built by OpenAI and Anthropic, Meta argues open weight systems distribute power to individual users rather than concentrating it among governments or large institutions.

Open weight models differ from truly open source software because they do not make all underlying code publicly viewable. Competitors like OpenAI and Anthropic largely rely on closed models, which function as black boxes that users cannot inspect. Zuckerberg criticized labs that build AI mainly for companies and governments, arguing that this approach shifts power toward institutions instead of individuals.

Glimmer, Meta's newly announced open weight model, was developed partly from Muse Spark, a closed source model the company introduced in April and updated last week. The system uses less computing power than rival models but remains strong enough to run AI agents, or software that completes tasks without human input, Meta said in a blog post. Chinese developers including Alibaba, DeepSeek and Moonshot have released advanced, more affordable open models in recent months, pushing American firms to cut prices and build cheaper alternatives.

Anthropic and OpenAI both hold lucrative government contracts, a point Zuckerberg raised while making his case for individual control over AI. "Rather than centralizing superintelligence, we should distribute it widely and give every person the ability to direct it," he wrote in the essay.

What legal challenges is Meta facing?

Meta's AI push comes as the company battles mounting legal and regulatory pressure at home. Meta has 3.6 billion users across Facebook, Instagram and WhatsApp, and concern over youth usage of its apps has grown sharply. A New Mexico court recently ordered Meta to pay nearly $1 billion in damages, and jury selection begins Wednesday in a California case brought by four states seeking $1.4 trillion in damages. The company still faces thousands of potential lawsuits nationwide over alleged harm to children who use its platforms.

How is the US government regulating AI?

Zuckerberg's essay and Meta's new models arrive amid a broader US debate over how to regulate AI developers, alongside growing local opposition to new data centers. President Donald Trump ordered a voluntary review process for advanced AI models in June, though his administration missed its own August 1 deadline to release details. Zuckerberg said close, proactive collaboration between frontier labs and the government would work better than a rigid review process in most cases.

He argued a flexible US regulatory environment for open models would give Meta an advantage over OpenAI and Anthropic and help it compete with Chinese developers. "Our goal should be for American open source models to be the best globally. This requires removing the hurdles that make it harder for American open source models to compete," Zuckerberg wrote. He added that even a one month delay could add significant risk to American leadership while letting foreign models advance.

How much is Meta spending on AI?

Meta plans to launch a $1 billion fund to directly support communities near its data center sites, a company spokesperson told AFP, as public opposition to construction grows. The company expects to spend up to $145 billion this year, nearly double its 2025 total, as it invests heavily in AI infrastructure and talent.