Zuckerberg Lobbied Trump Directly Against a National AI Regulator

Mark Zuckerberg called President Trump during the week of August 17 to voice opposition to a proposed national AI regulatory body — a rare direct lobbying moment that illustrates how one-on-one conversations between Trump and tech titans have become an unofficial backdoor into U.S. AI policymaking. According to a senior White House official, Trump actually initiated the call. Zuckerberg stopped short of demanding Trump kill the proposal, but urged that any appointees to such a body reflect Trump's own light-touch approach to AI regulation.

The regulatory concept being debated inside the White House is modeled on FINRA — the Financial Industry Regulatory Authority — and was popularized by Nobel Prize-winning Google DeepMind scientist Demis Hassabis, who pitched it in a July essay and briefed White House officials on it this summer. The idea: an industry-funded, independent body that would vet and stress-test advanced AI models for cybersecurity and other risks before they're deployed publicly. Proponents argue it could attract technical talent and create common evaluation standards. Critics worry voluntary standards could eventually harden into mandatory law.

Zuckerberg isn't alone in his skepticism. David Sacks — Trump's former AI and crypto czar — has mockingly called the FINRA model a 'DMV for AI' where models sit in a queue waiting for bureaucratic sign-off. Sacks is instead championing a voluntary ratings system modeled on the Motion Picture Association, which he says has backing from Elon Musk. On his All-In podcast on August 21, Sacks argued the MPA model succeeded precisely because voluntary industry standards preempted heavier government intervention. Sacks previously demonstrated his clout in May, when a single phone call to Trump the morning of a planned AI executive order signing convinced the president to cancel it entirely.

The broader tension reflects a real policy dilemma: AI agents are increasingly autonomous — capable of browsing the web, writing code, and operating software with minimal human oversight — and have already escaped human control in several high-profile incidents. A hostile actor wielding these tools could automate cyberattacks faster than existing defenses can respond. The White House says it's weighing two paths: the FINRA-style regulator or the voluntary MPA-style group. Zuckerberg's call did not kill the FINRA proposal, which remains under active consideration. The biggest AI labs — OpenAI, Anthropic, Google — have not publicly staked out positions, though Anthropic co-founder Jack Clark posted favorably about the concept in July.