Meta on Monday released Muse Glimmer, a small open-weight artificial-intelligence model designed to run on a single graphics card, and Chief Executive Mark Zuckerberg published a 14-page essay arguing that the greatest danger from powerful AI is not the technology itself but the prospect of one government or entity controlling it.
The release, paired with the essay "The Future is for Everyone," is Meta's clearest attempt to reset its AI strategy after last year's disappointing Llama 4 launch and to press Washington to loosen constraints on open models as Chinese labs pull ahead in that segment. Meta shares, down about 10 percent for the year, rose nearly 3 percent in premarket trading.
What Meta released
Muse Glimmer is smaller than the frontier systems built by OpenAI, Anthropic and Alphabet's Google, and is designed for agentic tasks that can run on a Mac or PC with a single graphics card, aimed at demand for AI that operates directly on user devices. Meta said it will publish the weights of Muse Spark 1.2, its most advanced model, in the coming weeks. "We've got even bigger models that are coming soon," Zuckerberg said in a video accompanying the essay.
The manifesto
Zuckerberg told Axios' Mike Allen that "I'm personally more worried about centralization than I am about any of the specific risks others are talking about." He added: "One of the ways to achieve both individual empowerment and checks and balances is to make sure you put the technology in everybody's hands." The essay warns: "Any policy that slows American model releases — even by a month — could add significant risk to American leadership while letting foreign models race ahead," a reference to Chinese open-weight systems such as Moonshot's Kimi K3, Alibaba's Qwen3.8-Max and DeepSeek's V4-Flash.
Policy and governance
Zuckerberg called on Washington to loosen restrictions on training data and model distillation, the practice of using a powerful model to train a smaller one. "Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data," he said. The Trump administration told developers earlier this month that it will not subject open-weight models to voluntary safety tests, Reuters reported, citing two people familiar with the discussions. Meta said its independent directors will gain the power to approve safety criteria for future model releases.
The community fund
Meta paired the launch with a $1 billion Future is for Everyone Fund to support teachers, first responders and energy and water infrastructure in communities that host its data centers, a response to local opposition that has become a hurdle to the company's roughly $145 billion in AI infrastructure spending this year.
AI safety researchers, dissenting industry executives and critics of Meta's data-use practices were not represented in the wire and center-tier reporting available for this article. Meta's current policies allow the company to use nearly every interaction with Meta AI to train its systems, Axios reported, though it recently added an incognito mode.
Meta said the weights for Muse Spark 1.2 will follow in the coming weeks, and America's Workforce Academy, its trades program with guaranteed jobs at company data centers, graduated its first class last week.