Nvidia, Meta CEOs Questioned Anthropic’s Dario Amodei Over AI Warnings: WSJ
AI industry leaders presented a united front at the White House this week, agreeing to a set of principles aimed at making artificial intelligence safer. But behind the scenes, some of the biggest names in the industry r
লেখক Ayush Mukherjee··12K বার পড়া
Nvidia, Meta CEOs Questioned Anthropic’s Dario Amodei Over AI Warnings: WSJ
AI industry leaders presented a united front at the White House this week, agreeing to a set of principles aimed at making artificial intelligence safer. But behind the scenes, some of the biggest names in the industry reportedly challenged Anthropic CEO Dario Amodei over his increasingly vocal warnings about the risks posed by advanced AI models.
According to The Wall Street Journal, Nvidia CEO Jensen Huang and other technology executives questioned Amodei after a White House lunch attended by around two dozen AI CEOs. The executives reportedly asked why Amodei had taken such a strong public position on the capabilities of AI models and the risks they could create.
Amodei defended his approach, arguing that the industry needs to be transparent about what increasingly capable AI systems can do rather than downplaying potential risks. The conversation took place as executives and White House officials were working on a set of principles for AI development and safety.
AI CEOs Agree On Safety Principles Despite Internal Differences
The discussions highlight a growing divide within the AI industry over how much emphasis should be placed on safety measures as companies race to develop increasingly powerful models.
The CEOs ultimately agreed to a set of principles after discussions involving US President Donald Trump and House Speaker Mike Johnson. However, the private questioning of Amodei suggests that there are still significant differences among leading AI companies over how the industry should approach AI risks and self-regulation.
The WSJ reported that these disagreements could complicate the industry’s efforts to establish a common framework for regulating itself through model evaluations, internal controls and other safety measures.
Amodei has previously warned about the possibility of AI being used to conduct cyberattacks and has raised concerns about potentially significant job losses as AI capabilities improve. He has also criticised both the White House and competing AI companies for what he sees as insufficient attention to safety concerns.
Despite those differences, Amodei struck a cooperative tone after the meeting. Speaking to reporters on Tuesday, he said, “we can win safely” if the AI industry and the US administration work together. Trump also praised Amodei following the discussions.
Jensen Huang And Mark Zuckerberg Push A Different Approach
Nvidia’s Huang and Meta CEO Mark Zuckerberg have reportedly emerged as prominent supporters of the White House’s more hands-off approach to AI development.
According to the WSJ, Zuckerberg played an important role in finalising the principles agreed upon Tuesday and held multiple discussions with Trump administration officials, including Commerce Secretary Howard Lutnick.
The report says Zuckerberg and Huang have pushed back against calls from researchers at Anthropic and OpenAI for the government to coordinate with the industry on potentially slowing AI model development if necessary.
The disagreement is not simply about whether AI should be made safer. Instead, the companies appear to differ over how that safety should be achieved, who should oversee it and whether government intervention should play a larger role.
Earlier this year, Anthropic, OpenAI and Google had reportedly supported an idea being considered by the administration for a self-regulatory organisation modelled somewhat on the Financial Industry Regulatory Authority.
That proposal was ultimately abandoned after Huang, Zuckerberg and SpaceX CEO Elon Musk reportedly expressed concerns that it could give too much power to the biggest AI companies.
AI Safety Debate Comes As Models Become More Capable
The latest disagreement comes amid increasing concerns about what advanced AI models could do as their capabilities expand.
The principles agreed upon Tuesday followed months of discussions focused partly on preventing AI-driven cyberattacks. The WSJ reported that the US administration has been testing powerful AI models before their public release, while recent incidents involving AI-assisted hacking have increased pressure for stronger safety reviews during model development.
For Anthropic, the debate also comes at a sensitive time. The company has been attempting to improve its relationship with the Trump administration after several disputes earlier in 2026.
Those clashes included a temporary shutdown of two Anthropic models over security concerns and a Pentagon decision to blacklist the company following a disagreement over the use of AI in military applications, according to the WSJ. Anthropic is also preparing for a potential initial public offering that could value the company at around $2 trillion, the report said.
AI Industry Wants Public Confidence
Despite the disagreements behind closed doors, the companies are publicly emphasising cooperation.
Meta’s Zuckerberg said the industry needs to give people confidence that AI systems operate as intended. He also argued that following the agreed principles collectively would be the way to address concerns raised during the discussions.
The contrast between the public agreement and the private debate illustrates the challenge facing the AI industry. Companies are competing to build increasingly capable models while simultaneously trying to establish common safety practices and reassure governments and the public that the technology can be developed responsibly.
The latest White House meeting therefore reflects two competing priorities within the AI sector: maintaining the pace of development and addressing concerns over what increasingly powerful AI systems could mean for cybersecurity, jobs and broader society.