Jensen Huang Tells Dreamforce: 'We Don't Need Any New Laws' as AI Leaders Publicly Split on Safety

The most consequential argument in artificial intelligence is no longer happening behind closed doors. It is happening onstage, in front of thousands of people, with the industry's most powerful executives openly contradicting each other. On Tuesday, Nvidia CEO Jensen Huang took the stage at Dreamforce in San Francisco — minutes after Anthropic CEO Dario Amodei had used the same stage to press his case for slowing down — and delivered the bluntest possible rebuttal: "We don't need any new laws. We don't need new regulations."

"Safety is an engineering problem, not a legal one," Huang told the audience, according to TechCrunch's September 15 report from the conference. The line drew a bright, public boundary through an industry that, only days earlier, had looked like it was converging on Amodei's call to "pace the frontier." It isn't converging. It is splitting — and the split was on full display again on Wednesday, when Meta CEO Mark Zuckerberg added his own dissent, telling Reuters that AI labs already have enough incentive to build safely on their own.

Dreamforce, Salesforce's flagship annual conference, drew some 12,000 attendees this year, according to Adweek — and it has become an unlikely arena for the defining policy fight in technology. The spectacle of Huang and Amodei appearing minutes apart, offering diametrically opposed theories of how AI should be governed, captures a moment in which Silicon Valley can no longer pretend to speak with one voice.

Nvidia CEO Jensen Huang told Dreamforce attendees that AI safety is "an engineering problem, not a legal one." (Photo: Peter Dasilva via the European Commission audiovisual service, CC BY 4.0 — attribution: © European Union)

Two keynotes, two philosophies

To appreciate how stark the split is, rewind four days. On September 12, Amodei published his essay "We Must Pace the Frontier," arguing that the AI industry should deliberately slow the improvement of its most capable models, grant independent evaluators employee-like access inside the leading labs, establish common safety standards among democratic nations, and eventually negotiate international limits on the most dangerous capabilities. The essay won swift endorsements from Elon Musk, OpenAI's Sam Altman, and Google DeepMind's Demis Hassabis — an extraordinary alignment of rivals that we covered in detail here.

Huang's answer, delivered to a Dreamforce audience that included many of the enterprise customers who buy his chips by the tens of thousands, was to reject the entire premise that new rules are needed. "So run as fast as you can," he said. "But if you feel at any given point in time the company's out of control, or the product's not going to be safe, you know, take a pause and make sure you get it right."

Note the structure of that argument, because it is doing a lot of work. Huang is not saying safety doesn't matter — he is saying it is a problem for engineers, product teams, and existing law, not for legislators writing new AI statutes. The pause he describes is voluntary and company-led: run fast, self-correct when needed. It is the classic Silicon Valley theory of governance — move fast, fix it yourself — stated at the exact moment a growing chorus of the industry's own leaders is declaring that theory insufficient.

And Huang is bullish to the point of exuberance about what the unfettered path delivers. "The sky's the limit for us… for every industry, for every single country," he said. For the CEO of the company whose graphics processors power the overwhelming majority of frontier AI training, that is both a forecast and a sales pitch: the AI build-out continues, the chips keep shipping, and the only speed limit is the one companies set for themselves.

Anthropic CEO Dario Amodei's call to "pace the frontier" won backing from Musk, Altman, and Hassabis — but not from Huang or Zuckerberg. (Photo: Kimberly White/Getty Images for TechCrunch via Wikimedia Commons, CC BY 2.0)

Huang's case: engineering, liability, and markets

Strip away the stagecraft, and Huang is making a substantive policy claim with three pillars. First, that AI safety is tractable as an engineering discipline — testing, red-teaming, guardrails, and alignment research — rather than something law can usefully specify in advance. Second, that existing legal frameworks, particularly around product liability and criminal law, already cover the harms that matter. Third, that market discipline does the rest: companies that ship unsafe products lose.

It is a coherent position, and it has a powerful constituency. President Donald Trump articulated a version of it on September 14, writing on Truth Social that "we already have tremendous CRIMINAL and REGULATORY power over these companies" and dismissing the slowdown campaign as a "HOAX." Trump is, notably, an ally of Huang's on this question: while Huang was onstage at a Los Angeles conference this week, the president called him mid-keynote, and Huang put the call on speakerphone as Trump declared "the robots will not be taking over" — to Huang's audible agreement.

But Huang's incentives deserve to be stated plainly, because they shape how the argument lands. Nvidia sells the picks and shovels of the AI gold rush. Any policy that slows frontier training — mandatory evaluations, capability caps, coordinated pauses — would, at the margin, slow demand for the chips that made Nvidia the most valuable company in the world. That doesn't make Huang wrong; the best argument against new AI laws might genuinely come from the man who best understands what the technology can do. But it does mean his "no new laws" stance aligns, with remarkable neatness, with his company's commercial interests. Readers should hold both facts in their heads at once.

There is also a genuine philosophical disagreement underneath the business logic. Amodei's camp argues that competitive pressure makes voluntary pauses unstable — that no lab can afford to slow down unilaterally while rivals sprint, which is precisely why coordination or regulation is needed. Huang's camp argues the opposite: that competition is the safety mechanism, because unsafe products get punished by customers, courts, and markets. These are not two wordings of the same idea. They are incompatible theories of how the most powerful technology of the century should be steered.

Zuckerberg breaks ranks too

If Huang's dissent was the headline act, Zuckerberg's was the confirmation that the split runs deep. Speaking on Wednesday, the Meta CEO told Reuters that competition and liability already give AI labs sufficient incentive to act independently on safety — no coordination required, no new mandates needed.

"Trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models," Zuckerberg said. "Any lab that doesn't focus on alignment will fall behind."

That is a revealing formulation. Where Amodei frames safety as a collective-action problem — everyone must slow down together or no one can — Zuckerberg reframes it as a competitive advantage: the labs that build the most trustworthy AI will win, and the market will sort the rest. It is, in miniature, the same argument Huang made at Dreamforce, translated from chips to models.

The practical upshot is a clean three-way map of the industry's power centres. On one side: Anthropic, OpenAI, xAI, and Google DeepMind — the frontier labs, now publicly committed to some version of coordinated pacing and independent evaluation. On the other: Nvidia and Meta — the infrastructure giant and the open-models giant — arguing that engineering, liability, and competition are enough. MarketWatch, in coverage published Wednesday, framed it as the defining split among Amodei, Altman, Huang, and Zuckerberg. It is hard to disagree.

Why this matters in Washington

The Dreamforce split is not academic, because Washington is moving — and the industry's disunity changes what legislation is possible. House Speaker Mike Johnson said this week he expects Trump to meet with AI executives within days to discuss "the responsibilities of the companies to maintain safety and what role, if any, the government has to play." Three senators — John Thune, Ted Cruz, and Amy Klobuchar — are working on a bill that would require AI companies to demonstrate "reasonable precautions" against harm, according to Reuters. OpenAI is backing separate legislation, the FRONTIER Act, that would mandate independent safety evaluations. And Vice President JD Vance has voiced open skepticism about executives "begging the government to regulate them."

Into that churn, the Huang-Zuckerberg position lands as a powerful counterweight. A Congress considering new AI mandates now faces not just industry opposition in the abstract, but named, specific, on-the-record dissent from the CEOs of two of the most valuable companies on Earth — dissent delivered on public stages, in quotable lines. "We don't need any new laws" is a sentence lobbyists can print on a placard.

It also complicates the politics for the slowdown camp. Amodei's coalition looked, for a brief moment last weekend, like the industry speaking with one voice — four rival lab leaders behind a single essay. Huang and Zuckerberg have now broken that image, and in doing so they have given political cover to everyone in Washington who preferred the status quo anyway. The Trump White House, which has framed AI dominance over China as the overriding priority, will find Huang's "run as fast as you can" far more congenial than Amodei's "pace the frontier."

That is the deeper significance of what happened at Dreamforce. The AI safety debate spent years as a conversation among researchers, then months as a lobbying battle, then a week as a public essay war. It has now become something else: an open, public schism among the people building the technology, playing out in real time, in front of customers, cameras, and lawmakers. There is no longer a single "industry position" on AI safety to report. There are two — and they cannot both be right about the most important technology of our time.

Sources: TechCrunch (Sept. 15, 2026); Reuters (Sept. 16, 2026); Adweek (Sept. 2026); MarketWatch (Sept. 16, 2026); TradingView/Reuters summary (Sept. 16, 2026); TechPulse USA reporting on the Amodei essay and Trump's response (Sept. 2026).