Nvidia and Meta CEOs Reject Coordinated AI Safety Slowdown

The leaders of Nvidia and Meta have publicly rejected calls from rival AI labs to coordinate a slowdown in research in the name of safety, arguing instead that individual companies should bear the responsibility for managing the technology’s risks. Jensen Huang, chief executive of the world’s largest supplier of AI hardware, and Mark Zuckerberg, who runs a major AI lab inside Meta, both distanced themselves on Tuesday from proposals advanced over the weekend by the heads of OpenAI, Anthropic, and SpaceX. That rare moment of unity among the three competitors came after mounting public concern and researcher warnings that an all-out race to advance AI could lead to humans losing control of the technology with catastrophic consequences.

Huang went further than Zuckerberg, arguing there is no need for new US laws or regulations on AI safety and urging labs not to fall into what he called a “false choice” between the speed of innovation and the ability to develop safe products. Speaking at an event with Salesforce chief executive Marc Benioff in San Francisco, Huang said market forces already provide sufficient discipline. “If you build a product or a service and you’re not confident in its functionality, capability or safety, then don’t release it,” he said. “That’s a very obvious thing to do: You pace yourself until you are confident you’re releasing something that the market will appreciate.” He added that companies could have both innovation and safety at the same time, advising labs to “run as fast as you can” but to pause if products appear unsafe.

The split between the two camps reflects deeper divisions within the AI industry over the future shape of the technology. The weekend’s signatories, including Anthropic’s Dario Amodei, OpenAI’s Sam Altman, and Elon Musk, lead developers of proprietary “closed” models that are tightly controlled and sold as services. Nvidia and Meta, by contrast, have advocated for more customizable “open” models that can be freely adapted and distributed. Huang has pushed the importance of open models to encourage widespread AI adoption, while Meta has released some closed models but generally argued for open systems. Some advocates of open models contend that regulations or controls on development would advantage the big labs that sell closed systems, entrenching their market position at the expense of smaller players and academic researchers.

The public disagreement also highlights the commercial stakes involved. Nvidia is a major investor in OpenAI, Anthropic, and Musk’s SpaceX, giving it a financial interest across the spectrum of AI development even as its hardware powers the training of virtually all leading models. Meta, for its part, has made open models central to its strategy, releasing them free of charge in an effort to shape industry standards and attract developers to its platforms. Google and Microsoft, which have signaled support for the coordinated slowdown proposals, are themselves deeply invested in proprietary AI systems, suggesting that the public positioning on safety aligns closely with each company’s competitive interests.

The policy implications are significant. Huang’s rejection of new US regulations comes as governments worldwide weigh how to govern a technology that is advancing faster than the legal frameworks designed to contain it. His alignment with former President Donald Trump, who took a phone call from Huang on stage at an event on Monday and has repeatedly rejected AI slowdown efforts, underscores the political dimension of the debate. The question of whether safety is best achieved through industry coordination, individual corporate discipline, or government regulation remains unresolved, and the divisions among the technology’s most powerful players suggest no easy consensus is forthcoming.

For investors and policymakers, the takeaway is that the AI industry is not monolithic. The public split over how to manage existential risk is, at its core, a battle over who controls the technology’s future and who profits from it. Until the underlying economics of open versus closed models shift, the industry is likely to remain fragmented, with safety proposals advancing unevenly and enforcement left largely to individual companies’ judgment.