The Wall Street Journal reports that Mark Zuckerberg, Jensen Huang and Elon Musk spoke directly with President Trump as a proposed federal AI oversight framework lost momentum. The account relies on people familiar with the discussions, so the conversations and their effect should be treated as reported rather than independently established.

The proposal, reportedly advocated by Google DeepMind chief executive Demis Hassabis, would have created a FINRA-style approach to oversight and pre-release testing for powerful frontier AI models. FINRA oversees brokerage firms and exchange markets in the United States. Applied to AI, the comparison suggests independent checks before the most capable systems reach the public.

According to the Wall Street Journal, Zuckerberg, Huang and Musk raised concerns directly with President Trump. The proposal then lost momentum. That sequence is important, but it does not by itself establish every participant's motive or prove that those conversations were the only reason the plan stalled.

The report illustrates how AI governance is being shaped through both formal policy processes and private discussions between political and industry leaders. Supporters of mandatory testing argue that increasingly capable models need independent scrutiny. Industry leaders may question whether a new oversight body would slow development, duplicate existing work or create rules before the technology is well understood.

The proposal was not a final law or settled government policy, and the reporting is based on unnamed sources. Its exact provisions, the administration's full reasoning and whether a revised version will return remain uncertain. Those limits matter because a stalled proposal is not the same as a permanent rejection of AI oversight.

For businesses and developers, the practical point is not to treat any single proposal as the final shape of US AI regulation. The more useful signal will be whether the administration publishes a revised framework, names an oversight body or sets specific testing expectations for frontier models.