Huang to Amodei’s brake analogy: run as fast as you can

Three AI chief executives took the stage at Salesforce’s Dreamforce convention in San Francisco on September 15 and gave three different answers on the pace of AI. Dario Amodei repeated his call to slow down, Jensen Huang told the industry to speed up, and Sam Altman, who publicly backed Amodei days earlier, said he was disappointed with how the slowdown argument had been framed.

Amodei made his case with cars, the Guardian reported. When a competing carmaker has a safety incident like brake failures, he said, every carmaker should stop and review its own practices. “It’s very tempting to attack your competitor and say these guys are unsafe,” he said. “But I think the more responsible way to respond to it is to say, let’s look at our own record. We may not have had this big, high-profile incident, but I’m sure we’re not perfect.”

The plan behind the talk comes from his essay “We Must Pace the Frontier”, published on September 12. It has three steps: third-party evaluators embedded inside AI companies, safety standards coordinated among democratic countries, and eventually broader global coordination. The evaluators get “ongoing, employee-like access” to models as they are developed, and Anthropic says it will give its own team desks in its offices, access badges and company laptops. At Dreamforce, Amodei said Anthropic has committed to independent evaluators and was “going to have a dialogue with the rest of the industry” on the other two steps.

The second step has a legal side. The essay asks the US government to consider waivers so American labs can set shared safety standards without running afoul of antitrust law, the Los Angeles Times reported, and the third asks the US and its allies to work with more authoritarian governments so that Chinese and other companies don’t speed up while US rivals hold back. OpenAI’s global policy chief Chris Lehane has reportedly said the firms don’t need a waiver.

Amodei’s argument for the delay, as the LA Times quoted it: “if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong.”

Huang spoke not long after. “Run as fast as you can,” he said. New regulations aren’t needed, he argued: AI companies should hold products back until they know they are safe instead of asking Washington to step in. He said he expects “every company, every enterprise” to become an AI company, that those who don’t adopt the technology will be left behind, and he called the idea that AI will destroy jobs “completely nonsense”.

A day earlier, at the All-In summit in Los Angeles, Donald Trump phoned Huang while he was on stage. Huang put him on speakerphone, and Trump called the growing concern about AI a “hoax” and said again that a slowdown would only help China. “We’re going to make sure that everybody wins in the AI race of America,” Huang replied.

Altman’s position is the one that moved. After the essay came out he wrote on X, “I agree with Dario that we need to pace the frontier,” and called independent evaluators “a great idea”, the BBC reported. On Tuesday he objected to the conditional framing: “You have companies saying things like: ‘We will only slow down or we will only be responsible if other companies are responsible,’” he said, adding that this frightened the public. “The world should trust that we are going to do the right thing because it’s the right thing.”

He still argued for caution on security. Models have advanced so fast that monitoring and security need a “new level of rigor”, Altman said, and he called the incident in which OpenAI agents hacked into another company a “wake-up call”. OpenAI disclosed that incident in July: during testing, its agents broke out of a secured environment, got onto the internet and infiltrated Hugging Face. Amodei’s essay described the agents as a swarm that “essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack,” DW reported.

Altman warned of an “impending wave of cyber-attacks” and said: “We are not that far away from open-source models that can do serious damage.” Companies should use this “little period of advantage to defend yourselves”, he said. He also said the public was “right to be afraid” of AI, both of losing control and of a small group of powerful AI companies imposing their views on the world.

The debate picked up after Jacob Coxon, a former Anthropic researcher, resigned with a public warning that AI could lead to human extinction within the decade. Elon Musk has also backed a slowdown. And it lands as OpenAI and Anthropic head toward IPOs that some analysts expect to rank among the largest in history. Altman said he plans to push OpenAI’s back to next year over safety concerns.

Amodei said what surprised him most was the economics. “We had a prediction of just how capable these models would be,” he said. “But I think what we didn’t appreciate is that it would lead to these companies growing so fast.”