Meta chief executive Mark Zuckerberg raised objections to a proposed federal AI regulator during an August phone call with President Donald Trump. The plan, championed by Google DeepMind co-founder Demis Hassabis, would create an industry-led body modelled on the securities industry's self-regulator.
A senior White House official familiar with the conversation said Zuckerberg told Trump the week of August 17 that he was against the idea. A second person familiar said Zuckerberg did not urge Trump to drop the plan, but suggested that anyone chosen to lead such a body should reflect Trump's preference for a light-touch approach. Trump initiated the call, that person added.
What happened
- Zuckerberg objected to a FINRA-style AI regulator in a call with Trump the week of August 17.
- The idea was popularised by Google DeepMind co-founder Demis Hassabis.
- The proposed body would examine advanced models for emerging risks before wider rollout.
- White House officials previewed the concept with Trump and major AI companies in mid-August.
- The call did not kill the proposal; officials are still weighing it.
Background: what FINRA actually does
The Financial Industry Regulatory Authority is a private, industry-led body that operates under SEC oversight. It writes and enforces rules for more than 3,000 brokerage firms and 630,000 registered representatives. Member firms fund its operations, and any rule changes require SEC sign-off.
The White House version would go further than rule-writing. Officials envision the new organisation scrutinising advanced AI models and probing them for cybersecurity and other emerging risks before they reach the market. Hassabis sketched the idea in a July essay, calling for a US "standards body" modelled on that approach, then walked senior administration officials through the concept this summer.
Our analysis: who gains and who loses
Winners: frontier model labs that want to keep release velocity high. Companies aligned with the administration's light-touch posture. Meta, which has now staked out a clear position with direct presidential access.
Losers: safety-focused researchers who want pre-deployment testing baked into the regulatory floor. Smaller AI startups that cannot absorb a parallel review process if rules land hard.
The strategic tension runs deeper than industry versus government. Hassabis argues that structured pre-release review reduces tail risk. Zuckerberg argued in a recent essay that slowing a model's release by even a single month could expose the United States to greater geopolitical risk relative to China. Both sides are selling fear of a different future.
There is also a leadership lesson here. Direct outreach from tech executives to the president has become a back-channel that can revive dormant ideas or sideline proposals that senior officials thought were settled. The substance matters, but so does knowing who can pick up the phone.
What it means for you
This is policy being shaped the way it usually is in Washington: in private, after the headlines move on. If you ship an AI product, treat the eventual shape of any federal AI oversight as a top-tier business question, not a back-burner compliance item.
For product and project leaders, the takeaway is timing. By the time a regulator announces itself, the window for design input is already small. Run scenario planning now. Map three possibilities: a FINRA-style body that pre-approves models for high-risk uses, sector regulators extending their existing remit, or voluntary industry standards only. The goal is not to predict the outcome. It is to know your exposure before the call lands.
Bottom line
The proposal is alive, the lobbying is active, and the August call did not settle it. Watch whether Hassabis's concept turns into a formal White House announcement before year-end. Watch whether Meta, OpenAI and Anthropic stay aligned in their objections or fracture. And watch whether "FINRA-style" is the final shape, or just the opening move.
Key takeaways
- Zuckerberg told Trump he opposes a FINRA-style AI regulator in an August call.
- Hassabis's concept remains under active White House consideration.
- Treat any AI regulatory shape as a top-tier business question, not a footnote.
Frequently asked questions
What did Mark Zuckerberg say to President Trump about AI regulation?
Zuckerberg told Trump the week of August 17 that he was against a FINRA-style AI regulator and suggested appointees should reflect Trump's light-touch approach.
Who proposed the FINRA-style AI regulator?
Google DeepMind co-founder Demis Hassabis floated the concept in a July essay and briefed White House officials on it this summer.
What would a FINRA-style AI body actually do?
Under the White House plan, the body would review advanced AI models and test them for potential risks before they go into broad use.
Did Zuckerberg's call kill the proposal?
No. The call did not quash the idea, and officials continue to weigh it.
What is FINRA, for context?
FINRA is an industry-led body that oversees more than 3,000 broker-dealer firms under SEC supervision, with member firms funding the organisation.
Source: businessinsider.com