Introduction
For years, the artificial intelligence industry has been embroiled in a heated debate over whether AI should be open – allowing anyone to tweak and tailor its algorithms – or closed, as a tightly held proprietary technology. Meta has now thrown its considerable weight behind one side, arguing that AI should be open and accessible to all
With the introduction of a new open-weight model called Glimmer plus an extensive public manifesto from CEO Mark Zuckerberg, Meta has positioned itself as a leader in the emerging fight over the future of artificial intelligence, with implications for both the technology and the regulatory environments in Washington and Silicon Valley.
What Is Glimmer, Meta’s New Open-Weight AI Model?
Glimmer is the latest in a series of open models from Meta – in this case, a set of parameters for building an AI that can be customized and deployed on individual computers, rather than in large data centers. As such, it is a much more modest and accessible take on frontier-scale AI – which is likely to put it in competition with similar efforts from OpenAI, Google, and Anthropic. Meanwhile, Zuckerberg’s public statement about the release of Glimmer provides insight into the company’s broader views about the technology and its regulation.
Zuckerberg’s Manifesto About the Future of AI
Along with the introduction of Glimmer, Zuckerberg published an essay articulating his thinking on the future of AI. His argument is straightforward: that advanced or “super” AI should not be centralized in the hands of a few firms but should instead be decentralized and distributed so that individuals and organizations can deploy AI in a manner best suited to their needs. Zuckerberg envisions a world in which every person has an AI assistant, an AI-enabled business tool, and an AI tutor at their disposal. He also argues that centralized systems are unable to account for the differing priorities and desires of society as a whole.
The timing of Zuckerberg’s remarks is also significant, as he faces growing criticism from both the public and policymakers regarding the massive energy costs of Meta’s new data center projects and the privacy implications of its upcoming smart glasses, which can allegedly record video of other people without their consent. In effect, Zuckerberg’s manifesto about the future of AI is also a response to criticism of his company’s approach to the development of the technology.
Why the Debate About Open vs. Closed AI Matters
The discussion around the merits and drawbacks of open versus closed AI is not just an abstract intellectual exercise. It has real-world implications for the future of the technology and the companies that seek to profit from it. For instance, there are some practical advantages to closed models: they are often more refined and safer, easier to monitor and test, and generate revenue for the companies that produce them.
Conversely, open-weight models allow companies to create products and services that build on top of existing AI, enabling broader adoption and use cases across a wide range of industries. They also offer developers the opportunity to innovate, with the strongest open models acting as foundational infrastructure upon which others can build. Furthermore, recent experience has shown that both open and closed systems can malfunction and engage in unacceptable behavior – meaning that neither approach is inherently safer or riskier.
Meta’s Strategic Position on Open vs. Closed AI
By choosing to champion open-weight AI, Meta has also signaled that it wants to appeal to a different set of customers than those that it has traditionally targeted. Until recently, Meta was one of the biggest proponents of open models in the industry, publishing a series of foundational large language models called Llama. However, late last year, the company announced that it would pivot toward a closed system called Spark, which it hopes will rival the models produced by OpenAI and Anthropic. By announcing the launch of Glimmer, a much more accessible open model, Meta appears to be attempting to straddle both camps.
Why Nvidia and Other Companies Support Open Models and What It Means for the Tech Industry
Nvidia CEO Jensen Huang has also been one of the strongest public proponents of open models – and he is not alone. The company has effectively marketed toward other firms looking to develop new applications and features enabled by open models, arguing that they provide a much-needed shot of innovation for the broader economy. Huang points to the use of open AI models by companies like Uber and Zoom in the creation of new products and services and their ability to analyze data more effectively. Meta, Google, and other companies have also pointed to the potential for open models to serve as critical infrastructure for the global economy. At the same time, many of the largest and most sophisticated open models in use today are produced by Chinese firms like Moonshot and DeepSeek, which has led to tensions in Washington, which has accused Beijing of attempting to steal trade secrets.
The Trade-Offs of Open Models and Why Some People Are Not Convinced That More Access Equals Innovation
There is no doubt that open models offer many advantages over their closed counterparts. However, many executives and industry insiders argue that, in practice, open models tend to lag significantly behind the most advanced closed systems in both capability and performance. Zuckerberg and Huang would certainly agree that having more companies participate in the development of AI will ultimately lead to broad gains for the economy and society as a whole, even if individual open models are not necessarily better than the most advanced closed ones.
Public Sentiment Toward AI and Why Meta and Other Companies Are Growing More Concerned About Regulation
Meta’s recent statements about its views on AI are also part of a broader pattern in which both the public and policymakers have grown increasingly hostile to the rapid rise of the technology and the stranglehold that a few gigantic firms have over its development. College students have booed executives who spoke about AI at graduation ceremonies; local governments have blocked the construction of data centers for generative AI from starting or completed construction; and Meta’s new smart glasses have been heavily criticized as invasive and creepy. Zuckerberg has also advocated for increased government oversight of the industry, arguing that collaboration between AI researchers and regulators was needed to establish a set of standardized, uniform safety protocols.
A public and political reaction of this magnitude would be difficult to imagine a decade ago, but the rise of frontier models has fundamentally changed the calculus. Zuckerberg has also said that regulation was appropriate given the extremely competitive nature of the industry – and a number of executives from Meta, Google, Anthropic, and OpenAI met with the White House this week to discuss the possibility of federal oversight. He argued that the government should have a role in reviewing critical models before they went live, a view that reflects the growing consensus among many tech executives that the AI industry has entered a “winner-take-all” era in which one dominant model will ultimately emerge.
What These Remarks Say About the Future of the AI Industry and the Regulation Debate in Washington
By publishing the Zuckerberg manifesto and open-sourcing Glimmer, Meta has signaled that it believes that competition in the AI industry will be fought over which companies get to dominate the market. However, the effectiveness of both strategies could be undermined if governments begin to treat open and closed models differently. Meta’s public statements about Glimmer should be seen as both an economic and a regulatory statement. It wants to be seen as a responsible company that encourages competition and innovation by championing open models, even as it invests heavily in a closed-weight model of its own called Spark that competes directly with OpenAI and Anthropic. For now, Zuckerberg has picked a side.
The next battle will be fought over whether open models can truly deliver on their promise to dramatically increase innovation and competition – or whether they will ultimately fail to meet the expectations of developers, businesses, and consumers. The debate over the future of artificial intelligence has only just begun, and it will likely have far-reaching implications for both the technology and the regulatory environments for years to come.
This article was previously published on CNN Business and was written ahead of Meta’s August 2026 announcement of the new open-weight AI model called Glimmer and the public release of a statement by CEO Mark Zuckerberg about the future of artificial intelligence.
