Meta CEO Mark Zuckerberg has publicly criticized the growing concentration of artificial intelligence (AI) development within a small number of dominant companies, while opposing proposed restrictions on Chinese AI models by the U.S. government. Speaking out amid escalating debates over AI governance and national security, Zuckerberg highlighted the benefits of broader access to AI technology and expressed skepticism toward approaches that prioritize tight controls.
In a recent series of interviews and a Wall Street Journal opinion piece, Zuckerberg argued that fears around AI’s potential to displace jobs and diminish human relevance have contributed to a doom-laden narrative that he believes is overblown. He suggested that centralizing AI power in the hands of a few companies—implicitly referencing firms like OpenAI and Anthropic—could stifle innovation and limit user access to the technology’s transformative potential. “The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous,” he wrote, warning against relying on a single “benevolent” superintelligence, which he described as unattainable given diverse human values.
Over the past weeks, the Trump administration has been finalizing an executive order that would encourage AI developers to submit new models for voluntary reviews lasting around 30 days, prompted in part by recent cybersecurity incidents involving AI systems. While Zuckerberg acknowledged the importance of addressing new challenges, he cautioned that such review periods, though appearing brief, represent significant time in a rapidly evolving field and should be implemented thoughtfully to avoid slowing innovation.
The U.S. government is also considering measures to limit access of American firms to Chinese AI models following allegations that Beijing-based Moonshot AI engaged in intellectual property theft to develop its Kimi K3 model, which has capabilities approaching those of leading American offerings. The Federal Communications Commission recently banned Chinese humanoid robots over national security concerns. Zuckerberg, however, questioned the efficacy of banning open-source models from other countries, suggesting that maintaining openness and accelerating domestic AI development would better serve American competitiveness.
Zuckerberg’s stance is part of a broader debate within the tech industry and regulatory circles. Leaders of Nvidia, Google DeepMind, OpenAI, Anthropic, and others have presented varying views on how to balance innovation with safety. Some, like Anthropic CEO Dario Amodei, advocate for stronger regulation and careful control of powerful AI models to mitigate risks. Others emphasize the value of openness and collaboration, with top AI researchers from major companies jointly calling for international efforts to implement safety standards and deliberate pacing of AI development.
Meta itself has shifted its approach recently. Once a champion of fully open-source AI models, the company has invested heavily in closed, proprietary AI research under the Meta Superintelligence Labs initiative, including the launch of its Muse Spark model as a commercial product. Meta also continues to support open-source projects but frames broader distribution of AI tools as essential to “individual empowerment” and democratization.
Alongside these policy discussions, Meta reported a record $60.8 billion in second-quarter revenue and plans to increase its AI-related spending by $5 billion, underscoring the company’s commitment to remaining a key player in the evolving AI landscape. The ongoing debate over how best to govern AI development reflects deep uncertainties about the technology’s future trajectory and its societal impacts, with Zuckerberg emphasizing the importance of multiple voices informing the discourse.