A series of private conversations between technology moguls and White House officials are significantly influencing the path of AI regulation in the US.

According to a senior White House official who spoke with Business Insider, Meta's CEO, Mark Zuckerberg, had a direct conversation with former President Donald Trump during the week of August 17th. In that call, he clearly stated his opposition to the proposal for a national-level AI regulatory body. The White House responded, stating the administration is committed to balancing innovation and safety in its AI policy approach. This disclosed phone call highlights the ongoing reality where top tech executives bypass standard policy channels to exert direct influence on the President. The regulatory proposal remains under internal White House discussion and has not been shelved due to Zuckerberg's objection. However, this episode underscores the persistent private lobbying pressure the industry exerts on the policy direction taking shape within the White House, pressure that can slow down or even redirect ongoing plans.

Proposal Origin: An AI Regulator Modeled on FINRA

The core proponent of this regulatory idea is Nobel laureate and Google DeepMind scientist Demis Hassabis. In an article published in July, he suggested that the US should establish a new AI 'standards body', similar to the financial industry regulator FINRA, to specifically address AI risks such as cybersecurity threats. A government official familiar with the matter revealed that Hassabis briefed White House officials on this concept during the summer. Google has not responded to requests for comment. FINRA is a private, non-profit membership organization operating under the oversight of the US Securities and Exchange Commission (SEC), responsible for creating and enforcing rules for over 3,000 brokerages and 630,000 registered representatives, funded by member fees with rule changes subject to SEC review.

According to White House officials, senior staff previewed this FINRA-like regulatory framework to Trump and major tech companies including Meta, OpenAI, and Anthropic in mid-August. The White House vision involves this agency reviewing and risk-testing AI models before they are officially deployed. Supporters argue this model can both attract the necessary technical talent for testing and establish unified standards for evaluating AI models.

Zuckerberg's Opposition: Aligning Oversight with Trump's Vision

Rather than voicing his stance publicly, Zuckerberg chose to communicate through private channels. The report, citing another person familiar with the matter, indicated that Zuckerberg did not ask Trump to change his position. Instead, he advised that if individuals were to be appointed to the regulatory body in the future, they should be selected for alignment with Trump's own AI philosophy – a stance widely perceived as favoring a lighter regulatory touch. The source also mentioned that Trump initiated the call. In a public article from August, Zuckerberg had already expressed reservations about strict government oversight. He argued that while government and AI companies should collaborate closely, any policy that could delay model releases – even by just a month – would pose a significant risk to America's leadership position relative to China. A Meta spokesperson declined to comment on the matter.

Internal Divisions: Two Paths Unresolved

This phone call sheds light on the deep divisions within the White House concerning the path for AI regulation. Officials are currently weighing two directions: one is establishing an industry regulator modeled on FINRA; the other is creating a voluntary industry group similar to the Motion Picture Association (MPA) – an approach championed by David Sacks, a former AI and cryptocurrency advisor to Trump. Sacks strongly opposes a government regulator, likening it to an 'AI version of the DMV' where models would be 'queued up for review'. Politico reported that in May, Sacks convinced Trump over a single phone call to cancel the signing of the first AI executive order on the day it was scheduled. On the 'All-In' podcast aired August 21st, Sacks detailed the plan he believes has Musk's support: an AI rating system modeled on the MPA, providing voluntary labels for AI models similar to film ratings. He noted that what the MPA did was promote standards, thereby preventing more intrusive, stronger government intervention. Musk has not responded to requests for comment.

Industry Cautious, Policy Direction Unclear

Among the leading AI companies, none have publicly endorsed or opposed the regulatory proposals. Anthropic co-founder Jack Clark posted on X in July, expressing some approval for the concept, but Anthropic declined to comment when asked about its official stance. OpenAI and Google have not responded to interview requests. Critics of the FINRA model have concerns, suggesting that voluntary standards and pre-market testing requirements, once established as practice, could eventually be incorporated into regulations and enforced by law.

This policy debate occurs at a critical point of rapid advancement in AI capabilities – AI agents can now autonomously browse the web, operate software, and write code with limited human oversight, and have already experienced loss of control in several high-profile incidents. In the hands of malicious actors, these systems could also identify vulnerabilities or launch cyber attacks faster than existing defenses can handle. The one-on-one meetings between Trump and tech executives have become an informal but highly significant variable in White House AI policy formation – one that could, at any time, slow down or redirect carefully laid plans, leaving some senior officials feeling constrained.