Zuckerberg Backs Nvidia's Huang in AI Safety Spat, Rejecting Calls to Slow Development
Meta's chief executive said he agrees with Nvidia's Jensen Huang that pausing or slowing AI work is the wrong response to safety concerns.
Meta chief executive Mark Zuckerberg has publicly aligned himself with Nvidia CEO Jensen Huang in the running argument over whether AI development should be slowed for safety reasons, according to comments reported by CNBC. The endorsement puts two of the most powerful figures in the AI supply chain on the same side of a debate that has split the industry for more than a year. Zuckerberg's position matters because Meta both builds frontier models and buys Nvidia's chips at enormous scale, giving the pair influence over how the argument is framed inside boardrooms and on Capitol Hill.
The disagreement has been simmering since early 2023, when thousands of researchers and executives signed an open letter calling for a six-month pause on training systems more powerful than OpenAI's GPT-4. That letter never produced a pause, and the companies that signed on to voluntary safety commitments in mid-2023 have since shipped successively larger models. Huang has argued that halting progress would hand advantages to competitors abroad and that safety work is better done while systems are being built, a view Zuckerberg echoed.
Zuckerberg's own track record gives the alignment extra weight. Meta has released its Llama family of models under open weights, letting outside developers download and modify them, a strategy that drew criticism from researchers who say it removes guardrails. The company has also stood up internal safety teams and published model cards, though it has disbanded or reorganized some of those groups during rounds of layoffs. Its stance is therefore not pure deregulation but a bet that openness and continued scaling produce better outcomes than restraint.
The practical stakes are large. Nvidia's data-center revenue has climbed into the tens of billions of dollars per quarter on demand from Meta, Microsoft, Google and Amazon, and any coordinated slowdown would hit that pipeline directly. Meta, meanwhile, has told investors it will spend heavily on AI infrastructure this year, with capital expenditure guidance in the tens of billions. Both companies have reason to keep the build-out moving, which is why their shared position draws scrutiny from safety advocates.
What to watch next is whether the Zuckerberg-Huang line hardens into a formal industry bloc or stays a loose alliance of convenience. Regulators in the European Union and several U.S. states are drafting rules that would impose testing and disclosure requirements on large models, and the outcome may depend on whether the largest labs present a united front. The open question is whether voluntary commitments, backed by the two most consequential companies in the stack, will be enough to head off binding restrictions.