Zuckerberg, Musk and Nvidia's Chief Just Helped Kill an AI Watchdog
The people building the most powerful AI just argued nobody new should police it.
Three of the most powerful people in technology — Mark Zuckerberg of Meta, Elon Musk of xAI, and Jensen Huang of chipmaker Nvidia — reportedly went straight to President Trump in recent weeks to argue against a new AI watchdog. The proposal came from Demis Hassabis, who runs Google's DeepMind AI unit. He wanted an independent body, paid for by the AI industry, that would test the most powerful AI systems before companies released them — checking for dangers like cyberattacks, bioweapons know-how, or AI that lies to get what it wants. He modeled it on FINRA, the private group that polices stockbrokers under government watch. Trump didn't agree to create it, reportedly frustrating some White House staff.
The pushback wasn't simply "companies hate rules." Zuckerberg, Musk and Huang reportedly worried the new body would hand extra influence to OpenAI, Anthropic and Google — the three labs already closest to the center of the AI race. If those companies help write the standards, competitors would have to follow rules shaped by their rivals. That's a legitimate concern, but it's also a convenient one for companies that would rather set their own pace.
Huang was the bluntest. "We don't need any new laws. We don't need new regulations," he said at a Salesforce conference. He argues safety is an engineering problem companies should solve themselves by holding back products they aren't confident in. Zuckerberg said AI firms already have strong incentives to be safe. Musk has been harder to pin down — he's mocked doom warnings but recently backed Anthropic's CEO when he called for more caution.
Meanwhile OpenAI's Sam Altman and others have backed stronger safety coordination. The Trump administration leans toward light regulation, often framing AI policy as a race against China — even as some officials still talk about AI risks. So for now, the answer to "who checks whether AI is safe?" is mostly: the companies themselves.
- The plan was for an industry-funded AI watchdog — think of it as a referee paid by the teams, similar to how FINRA polices stockbrokers.
- Zuckerberg, Musk and Huang argued it would concentrate power around OpenAI, Google and Anthropic, who'd help write the rules.
- Trump declined to set it up, so AI companies largely keep policing themselves for now.
Why It Matters
Self-policing AI means faster new tools for you — but fewer outside checks when something goes wrong.