Meta restarts open-source model strategy, Zuckerberg criticizes "closed-source" AI competitors

Meta has once again opened up some of its self-developed artificial intelligence models to external developers for free. The company's CEO, Mark Zuckerberg, stated that he opposes the monopolization of high-performance artificial intelligence technology by a few enterprises and governments.

The $1.5 trillion tech giant released the underlying parameters of its new open-source AI model, Muse Glimmer, on Monday, allowing developers to download and modify the model. Meta also revealed that it will release the underlying parameters of the more powerful Muse Spark version (referred to as "model weights" in the industry) in the coming weeks.

Accompanying this release, Zuckerberg published a column on Meta's official website outlining its strategy: to provide high-performance AI technology for free to billions of people, thereby empowering individuals while "balancing the power held by various institutions."

His remarks clearly target competitors such as Google, Anthropic, and OpenAI, stating, "The vast majority of competing labs only develop AI products for enterprises, governments, and various institutions. If the industry is led by such labs, the balance of power will significantly tilt towards large institutions, leaving ordinary individuals at a disadvantage."

This commitment marks Meta's return to its "open-source AI" strategy, which is a core differentiator from its competitors.

Earlier this year, competition among leading AI companies over the scope of top technology openness intensified. At that time, Meta had paused the release of the Muse Spark model weights, citing safety risks.

On Monday, Meta also announced the establishment of a $1 billion special fund to support communities where data centers are located in the United States, aiming to accelerate the infrastructure needed to support its AI development vision. The construction of data centers by major tech companies has sparked strong opposition from local residents, as data centers consume large amounts of electricity, freshwater, and other scarce public resources, leading to resource competition with local livelihoods.

Currently, Meta is investing hundreds of billions of dollars to tackle cutting-edge AI technologies, and Zuckerberg continues to promote his vision of "personal superintelligence": creating high-end intelligent assistants to comprehensively assist ordinary people with health management, hobbies, financial planning, career development, and other matters. This statement comes in the context of this industry background.

However, last month, Meta disclosed that its massive investment in AI infrastructure caused its free cash flow to plummet by 91%. Following this news, the company's stock price fell nearly 8%. Over the past year, this social media giant has spent heavily to attract top AI talent and build AI data centers, with its stock price dropping nearly 20% in total.

Zuckerberg stated that the company will continue to increase its investment in the development of open-source weight models and criticized competing labs for adopting a closed-source operating model—relying on selling model access rights to generate billions in revenue "Some opinions suggest that the optimal way to reduce the risks of artificial intelligence is to limit the technological capabilities accessible to ordinary individuals." Zuckerberg countered, "It has been proven that large-scale open-source systems are more secure because a vast number of developers can collaboratively identify system vulnerabilities." He cited the artificial intelligence startup Hugging Face as an example: the company had sought help from a closed-source model vendor to fix security vulnerabilities but was refused on security grounds, ultimately relying on open-source models to complete the system patch update