Anthropic CEO Dario Amodei (center) and Nvidia CEO Jensen Huang (front row, right) watch a press conference by House Speaker Mike Johnson (front row, left) after an AI safety meeting at the White House on Tuesday (local time). [AFP]
Anthropic CEO Dario Amodei (center) and Nvidia CEO Jensen Huang (front row, right) watch a press conference by House Speaker Mike Johnson (front row, left) after an AI safety meeting at the White House on Tuesday (local time). [AFP]

AI company leaders engaged in a fierce debate before White House staff over differing assessments of AI risks before signing a voluntary self-regulation agreement Tuesday (local time), according to people familiar with the matter. The sharpest divide emerged between Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg — both of whom enjoy cordial relations with the current administration — and Anthropic CEO Dario Amodei, who has had a fraught relationship with the Trump administration, underscoring how ties to the government appear to color executives' views on AI threats.

According to the Wall Street Journal, Huang confronted Amodei during a meeting with White House staff in the Roosevelt Room, held shortly after a White House luncheon, demanding to know why Amodei continued to issue such extreme public warnings about the dangers of AI models. Huang suggested Amodei was exaggerating the risks and stoking unnecessary alarm.

Amodei pushed back, telling the assembled AI executives that they had an obligation to be honest with the public about their models' capabilities and must not downplay the risks, the Journal reported.

Zuckerberg sided with Huang, the Journal said. During the luncheon, after Amodei raised safety concerns, Zuckerberg said the industry could address such worries by rigorously adhering to its own principles — signaling his opposition to new regulations.

Anthropic, OpenAI and Google had pushed to establish a self-regulatory body modeled on the Financial Industry Regulatory Authority to manage AI risks. But Huang, Zuckerberg and Elon Musk, CEO of SpaceX, told President Donald Trump they saw no need for a new body, and the debate ultimately converged on industry self-regulation as the preferred approach.

Amodei has repeatedly called for a slower pace of AI development, warning of its dangers, and a significant number of industry figures have agreed. Huang, however, has actively pushed back, arguing that speed and safety can be achieved simultaneously. Zuckerberg has also opposed external regulation, saying individual companies should set and maintain their own appropriate development pace without outside oversight.

The fault lines track closely with each company's relationship with the current administration, drawing considerable attention. Meta clashed with Trump after suspending his social media accounts following the Jan. 6, 2021, Capitol riot. The company has since settled related litigation and brought in figures with ties to Trump, rapidly repairing the relationship. The Journal reported that Meta has also made substantial donations to projects championed by Trump.

Anthropic, by contrast, has been at odds with the administration — unusually for a US company, the US Department of Defense designated it a supply chain risk. When tech leaders including Musk gathered for a state dinner welcoming Chinese President Xi Jinping on Sept. 24, Amodei was absent. More recently, signs of a thaw have emerged: Amodei dined with Trump at the White House on Sunday.


kate01@heraldcorp.com