A new battle over who controls advanced AI is intensifying, as Meta presents a model designed to move computing closer to everyday devices.
Meta has backed the view that artificial intelligence technologies should be available for users to customize, rather than being fully controlled by the companies that created them. In doing so, the corporation outlined its position in the debate between supporters of open and closed AI models.
The company introduced Glimmer, a new open-weight model. Its parameters can be modified to adapt the system’s behavior and responses to specific needs. Meta expects the model to become the foundation for AI tools that will run on personal computers without requiring a constant connection to data centers.
Mark Zuckerberg also published a manifesto explaining Meta’s artificial intelligence strategy. He addressed AI development, the construction of large data centers, and the company’s smart glasses.
Instead of centralizing superintelligence, we should distribute it widely and give every person the ability to control it.
– Mark Zuckerberg
Why Meta Is Betting on Open Weights
Closed models are typically controlled by their developers. Supporters of this approach believe it allows systems to be tested more thoroughly, dangerous use cases to be restricted, and a higher level of technological capability to be maintained.
At the same time, even closed AI systems can exhibit unpredictable behavior. Reports of cases in which models acted without adequate oversight are fueling debate over whether limited access truly guarantees safety.
Open-weight models allow other developers to use an established technological foundation for their own products and features. Nvidia CEO Jensen Huang also supports this approach. According to Nvidia, its open models are used by Uber and Zoom, including for developing new features and analyzing data.
Previously, Llama was one of Meta’s best-known open-weight models. At the end of last year, the company shifted its focus to its own closed model, Spark, which is seen as a competitor to the most powerful offerings from OpenAI and Anthropic.
At the same time, the rapid development of Moonshot and DeepSeek open models in China has intensified debate around Meta’s strategy. Some critics believe that restricting access to models could give Chinese developers an advantage, as they are more actively distributing their own technologies.
Opponents of the open approach emphasize that such models often lag behind their closed counterparts in terms of capabilities. Zuckerberg and Huang are convinced that wider distribution of the technology, on the contrary, will help accelerate the emergence of new solutions.
Artificial Intelligence Risks and Competition Between Companies
In his manifesto, Zuckerberg described artificial intelligence as a personal assistant, a tool for business growth, and a tutor with the knowledge level of “a PhD in every subject.” In his view, such important tools should not be concentrated under the control of only a few major corporations.
There is no technological solution that could simultaneously reconcile everyone’s opposing interests and values. Any single superintelligence would have to prioritize some values over others and therefore could not be benevolent toward everyone.
– Mark Zuckerberg
Alongside technological prospects, concerns are also growing about the consequences of rapid AI development. Key risks include potential effects on employment, cybersecurity, mental health, the economy, and the lives of communities near which data centers are being built.
According to Data Center Watch, projects worth approximately $130 billion in data centers were blocked or delayed in the first quarter of 2026 due to local opposition.
Zuckerberg said that Meta is building out its own energy infrastructure to prevent electricity costs from rising in the areas where its data centers will operate. Following a White House meeting involving representatives of Meta, OpenAI, Google, and Anthropic, he also called on leading AI labs to work more closely with the government when testing models before launch.
To develop the most advanced models and products, we must acknowledge that AI is likely the most competitive industry in history.
– Mark Zuckerberg
Meta is therefore seeking to combine the development of its own closed products with support for open AI models. The company believes that access to technology and the ability to adapt it will become key factors in future competition in the artificial intelligence market.