Report: The United States is considering establishing an independent AI regulatory agency to conduct safety reviews of top AI models.

```

After Silicon Valley tech leaders complained about the lack of consistency and transparency in recent U.S. government measures restricting the release of advanced AI systems, the Trump administration is considering establishing an independent regulatory agency with industry participation to conduct safety reviews of AI models.

According to Bloomberg, citing informed sources, Treasury Secretary Bessent has been involved in drafting a proposal to create an independent AI regulatory agency modeled after the Financial Industry Regulatory Authority (Finra), which would be accountable to the U.S. Securities and Exchange Commission (SEC). The proposal is currently under review by White House Chief of Staff Susie Wiles.

The background of this proposal is the public dissatisfaction expressed by Anthropic and OpenAI after government intervention in their model release processes.

Last month, Anthropic’s Fable 5 and Mythos 5 models were temporarily taken offline due to U.S. export control measures; OpenAI was required to make major modifications to its latest Sol model at the government's request. Both companies believe these actions are excessive relative to the actual safety risks.

Proposal Modeled After Finra Structure, Industry Participation in Standards Setting

According to informed sources, the proposal aims to establish an independent regulatory agency funded by the industry and operating within the SEC’s authorized framework—a structure closely akin to how Finra oversees brokers.

The agency would allow the financial and tech industries to jointly participate in setting AI safety standards, thereby strengthening cybersecurity controls while reducing the uncertainty impact on AI companies. This framework is still under discussion and details may change.

It is not yet clear what types of assessments the new agency would conduct on models, how it would be funded, or what role the SEC would play in regulating AI companies.

A White House official said the government is evaluating multiple plans to consolidate America's leading position in AI and strengthen cybersecurity.

DeepMind Proposal Gains Tech Industry Support, Aligns with Government Thinking

Earlier this week, Google DeepMind CEO Demis Hassabis released a set of policy recommendations highly consistent with the proposals currently under discussion by the U.S. government.

Hassabis recommended forming a committee of independent technical experts to review AI models within a standards agency framework supervised by the federal government. Required funding and computing resources would be provided by the industry, which would work with government agencies and U.S. national laboratories to develop model testing protocols for fields such as cybersecurity and biological threats.

In a post on social media platform X, Hassabis explicitly compared his envisioned regulatory body to Finra. The proposal has received public praise from Microsoft CEO Satya Nadella, OpenAI CEO Sam Altman, and SpaceX CEO Elon Musk—these three are typically seen as competitors to each other.

Hassabis will travel to Washington next week to lobby policymakers about this proposal.

As for the current regulatory situation, the Trump administration has long had internal disagreements over AI safety policy orientation. Anthropic’s Mythos model, which possesses cyberattack capabilities that have caused concern among officials and business circles, forced the government to reconsider the extent of its intervention.

The latest proposal is closer in direction to an executive order previously signed by Trump, which outlined a lightly interventionist, voluntary review mechanism jointly established with AI companies.

Risk Warning and DisclaimerThe market carries risks, and investments should be made cautiously. This article does not constitute individual investment advice, nor does it consider the investment objectives, financial circumstances, or needs of any individual user. Users should consider whether any opinions, viewpoints, or conclusions contained herein are appropriate to their particular circumstances. Investment decisions made based on this article are at your own risk. ```