Abstract:
On October 1, according to the Wall Street Journal, on Tuesday, after agreeing to sign a statement of principles aimed at ensuring the security of AI models, a number of AI company CEOs gathered around President Trump, demonstrating an industry image of unity in response to public concerns outside the White House.
Caption: Amodei
But according to people familiar with the matter, behind the scenes, industry leaders including Nvidia CEO Jensen Huang have just questioned Anthropic CEO Dario Amodei about his previous warnings about the dangers of AI.
After lunch with about 24 CEOs in the East Room of the White House, the executives asked Amodei why he spoke so publicly about the capabilities of AI models and the risks they pose. Some said Amodei told other executives it was important to be honest with the public about model capabilities and not downplay risks.
The conversations were among a smaller group of executives and White House aides and advisers who were hammering out AI principles in the Roosevelt Room, according to some people familiar with the matter.
Trump and House Speaker Mike Johnson pushed major AI companies to agree to AI principles at a lunch meeting, but the questioning of Amodei once again shows the undercurrent between Anthropic and other companies in the industry. This tension may hinder the industry's plan to use model evaluation and internal controls to achieve "self-regulation."
Amodei has previously criticized the White House and rivals for downplaying security concerns and warned that AI models could carry out cyberattacks and that the technology could lead to extreme unemployment, making him the target of criticism from many in Washington and Silicon Valley.
Amodeei told reporters on Tuesday that if the industry and the government work together, "we can win while ensuring safety." Trump praised him as "excellent." The two met for dinner at the White House last Sunday, their first face-to-face meeting.
Zuckerberg in particular played a key role in finalizing Tuesday's AI principles, some people familiar with the matter said, speaking frequently with Trump and others in the administration, including Commerce Secretary Howard Lutnick, who oversees a key government AI testing unit. He joined Huang in becoming a champion of the White House's light-touch regulatory approach and pushed back against claims by researchers at Anthropic and OpenAI to begin slowing down model development, if necessary, in coordination with the government.
According to some people familiar with the matter, when Amodei raised concerns about AI safety at the lunch meeting, Zuckerberg responded that the best way to solve these problems was for the entire industry to implement these principles.
Previously, Meta was criticized by Trump for suspending Trump’s social media account after the attack on the U.S. Capitol on January 6, 2021. The company has since aligned itself with the White House. During Trump’s second term, the company settled lawsuits over the matter, eliminated its fact-checking protocols, hired Trump-friendly executives and donated money to projects the president supported.
Comments