Huang And Zuckerberg Reject Calls To Slow Frontier AI Research

Both back a company-led approach to AI safety.

Nvidia chief Jensen Huang and Meta boss Mark Zuckerberg have rejected calls for technology companies to jointly slow the development of advanced AI systems.

The executives said individual AI companies should take responsibility for managing safety risks rather than agreeing to a collective pause in research.

Their comments came after Anthropic chief executive Dario Amodei, OpenAI's Sam Altman and SpaceX founder Elon Musk backed proposals for a coordinated approach to slowing frontier AI development.

Huang said there was no need for additional US laws or regulations specifically aimed at AI safety, arguing that market pressures already gave companies an incentive not to release unsafe products.

"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 at an event with Salesforce chief executive Marc Benioff in San Francisco.

He rejected what he described as a "false choice" between rapid innovation and safe AI development.

"You could definitely have both [innovation and safety] at the same time," Huang said.

He added that companies should slow down their own work if they believed a system was becoming difficult to control or was not safe enough to release.

Huang has previously dismissed warnings about AI posing an existential threat to humanity. He also appeared to support President Trump's opposition to proposals that could slow US AI development.

During the event on Monday, Huang took a call from Trump, who again described existential concerns about AI as a "hoax" and criticized efforts to restrict US research and datacenter construction.

Zuckerberg Backs Company-By-Company Approach

Zuckerberg similarly argued that AI labs should determine their own pace of development while putting safety measures in place.

In a post on X, he said every lab had both the responsibility and incentive to ensure its models were trained safely.

He pointed to Meta's own decision to delay its latest Muse model by several months because of internal safety concerns.

The Meta chief said the company had not asked other labs to slow down before taking action itself, arguing that safety decisions should form part of the normal development process.

He also called for greater use of independent evaluators and advisers to assess AI systems.

Zuckerberg suggested that a larger third-party AI evaluation industry could help companies check whether their systems were being developed safely and according to applicable rules.

AI industry Divided Over How To Manage Risks

The disagreement signals a wider divide within the AI industry over how the risks of increasingly capable systems should be managed.

Amodei, Altman and Musk recently called for greater co-ordination between leading AI laboratories, following warnings from researchers that an uncontrolled race to develop increasingly powerful systems could make it harder for humans to maintain control.

Google and Microsoft have indicated support for the broader idea of coordinating the pace of frontier AI development.

The debate also illustrates differences between companies developing proprietary, or "closed", AI models and those promoting more customizable "open" systems.

Meanwhile, AI regulation is also being discussed internationally.

Representatives from all 27 EU member states are due to meet in Brussels this week for a session of the bloc's Artificial Intelligence Board.

The European Commission is expected to provide an update on European and international AI policy, as well as a technical briefing on recent AI-related incidents.

This article originally appeared on MES Computing's sister site Computing.