The White House is weighing a mandatory FINRA-style AI review body against a voluntary MPA-style rating system, with no decision reached after Mark Zuckerberg privately urged President Trump to reject the heavier-handed option. The call, one of several direct lobbying channels shaping policy, pits a pre-deployment testing regime backed by Google DeepMind's Demis Hassabis against a lighter-touch model favored by adviser David Sacks.
The White House is weighing two competing models for governing frontier artificial intelligence — a mandatory FINRA-style review body or a voluntary industry rating system — with no decision reached weeks after Meta Platforms Inc. Chief Executive Officer Mark Zuckerberg privately urged President Donald Trump to reject the heavier-handed option.
Zuckerberg told Trump during a call the week of Aug. 17 that a national AI regulator was a flawed idea, according to two people familiar with the conversation cited by Business Insider. He did not ask Trump to abandon the proposal, the people said, but urged that any appointees to such a body share the president's light-touch view of AI oversight. Trump placed the call, they added. The White House said it remains "committed to balancing innovation and security" in AI policy.
The call is the latest instance of tech executives steering policy through direct access to the president rather than conventional channels, and it has exposed a divide inside the administration. Senior officials in mid-August briefed Trump and major labs — Meta, OpenAI and Anthropic — on a plan, championed by Nobel laureate and Google DeepMind chief Demis Hassabis, to create an AI "standards agency" modeled on the Financial Industry Regulatory Authority. FINRA, a private nonprofit under Securities and Exchange Commission oversight, polices more than 3,000 brokerages and 630,000 registered representatives, funded by member fees with rule changes subject to SEC review. The AI version would test frontier models for risk before release.
The alternative, pushed by former AI and crypto adviser David Sacks, would copy the Motion Picture Association's voluntary rating labels rather than create a government body. Sacks has likened a federal regulator to an "AI DMV" where models "queue for review," and on the Aug. 21 "All-In" podcast outlined a system, which he said Elon Musk supports, that would tag models with voluntary ratings akin to movie classifications. "What the MPA did was promote standards, which prevented more intrusive government intervention," Sacks said.
Sacks carries unusual weight with Trump: Politico reported that in May he persuaded the president by phone to cancel the first AI executive order on the day it was to be signed. Zuckerberg, for his part, wrote in August that any policy delaying model releases by even a month would pose "significant risk to American leadership" relative to China, aligning Meta's open-weight Llama strategy with the administration's deregulatory bent and setting it apart from rivals OpenAI and Anthropic, which favor more structured oversight.
No major AI company has publicly endorsed or opposed either path. Anthropic co-founder Jack Clark voiced support for the concept in July on X, though the company declined to comment on its official stance; OpenAI and Google did not respond to requests for comment. Critics warn that voluntary standards and pre-market testing, once established as practice, could harden into binding regulation.
The debate is unfolding as autonomous AI agents — able to browse the web, operate software and write code with limited human supervision — have lost control in several high-profile incidents and could, in hostile hands, identify vulnerabilities or launch cyberattacks faster than existing defenses can respond. That capability curve is what makes the choice consequential: a mandatory pre-deployment review could delay model launches and raise compliance costs across the sector, while a voluntary framework would leave testing to the industry. The administration has already ruled out any central licensing agency resembling an FDA for AI, and the European Union's AI Act, which imposes comprehensive obligations, stands as the counterexample if Washington opts for lighter rules. Officials have given no timeline for a decision, leaving the sector's regulatory burden — and the pace of American model releases — unresolved.
