Circle CEO disagrees with Huang on AI safety

Jeremy Allaire rejects treating the issue as civil liability and calls for public policy, in a video from The Information.

On 1 October 2026, The Information posted a 1 minute 23 second video on YouTube and on X. The X post went out at 18:16 in Brasília (21:16 UTC), under the title “Circle CEO Challenges Jensen on AI”. In it, Jeremy Allaire, cofounder and CEO of Circle, disagrees with Jensen Huang, CEO of Nvidia, on AI safety.

Circle issues USDC, a stablecoin: a digital token the company says is backed by dollar assets. Allaire does not run a frontier model lab and does not make chips. Huang sells the graphics processors used to train and run large models.

First: what the argument is about

Civil liability, in law, is the mechanism for who pays if a product goes wrong. Huang treats that, plus engineering and the market, as enough. Allaire does not accept that reading.

Alignment, in this field, is the work of getting a system to behave as people intend. Allaire says the labs at the frontier still lack a fundamental grasp of that. AI supercomputers, in his wording, are the large machines used to train and inspect those models.

What Allaire said on AI safety

The words are on the audio. “I don’t agree with Jensen’s view.” Labs at the frontier of these “AI supercomputers,” he said, “do not have an understanding of the like fundamental safety alignment of these systems.” Even when they try, “they’re having to use the AI supercomputers themselves to even attempt to understand the model safety and alignment.” He adds: “it feels very precarious.”

He also said: “I don’t agree that this is just a liability question.” And: “the stakes are extraordinarily high as well for society… We’re dealing with technologies that themselves are kind of weapons-grade systems.” With significant uncertainty and systems that “can be deployed in a variety of different ways,” that would be “a very real policy environment.”

“Weapons-grade” is Allaire’s analogy and opinion. It is not a technical or military classification. The clip has no figures and does not show that the models are weapons.

Huang’s line at Dreamforce

On 15 September 2026, at Dreamforce in San Francisco, Huang spoke with Marc Benioff of Salesforce. Politico and TechCrunch recorded: “We don’t need any new laws. We don’t need new regulations.” TechCrunch, in a 15 September 2026 piece by Julie Bort, also quotes: “Safety is an engineering problem, not a legal one.” And: “The market forces are already there.”

Huang said that if a company is not confident in a product’s functionality, capability, or safety, “then don’t release it.” Politico and TechCrunch both have that wording.

The day before, 14 September 2026, at the All-In Summit in Los Angeles, Donald Trump called Huang on speakerphone and labeled existential AI warnings a “hoax.” Huang, according to TechCrunch, said: “we’re not going to let that happen.”

Why a stablecoin CEO is talking about this

The disagreement comes from a digital payments company, not from a rival lab. Circle operates under financial regulation. That does not make Allaire a technical referee of frontier models.

The “weapons-grade” analogy points at dual use: civilian technology that can also serve military ends. It stops helping if you read the phrase as proof of a weapon. The video does not offer that proof. Allaire asks for a policy environment and does not describe which rule he wants.

What we still don’t know

We did not find the date, place, or format of the original interview, only the 1 minute 23 second cut. It is also unconfirmed whether Allaire was answering the Dreamforce remarks or some later comment from Huang. RUMOR: a written Circle statement or an Allaire post on X restating “weapons-grade.” RUMOR: a public reply from Huang or Nvidia to Allaire. RUMOR: a long open The Information article with the full recording.

Allaire disagrees with Huang in public, refuses to reduce AI safety to civil liability, and treats frontier systems, in his view, as something close to “weapons-grade.” If Huang replies, that is not yet on the record.

Sources

Press

Official

Circle, About

First: what the argument is about

What Allaire said on AI safety

Huang’s line at Dreamforce

Why a stablecoin CEO is talking about this

What we still don’t know

Sources

Sources

Markers