Silicon Valley is dissatisfied with the "sudden" export controls, the Trump administration plans to establish an "AI version of FINRA" to alleviate industry concerns.

According to informed sources, the Trump administration is considering setting up an independent regulatory agency to review the security of artificial intelligence models, with industry participation in the evaluation. This proposal comes after several leading figures in Silicon Valley complained that the recent practice of the US to temporarily suspend the release of cutting-edge AI systems was too "spontaneous." The sources indicated that Treasury Secretary Besent is assisting in drafting the proposal. According to the proposal, an independent AI regulatory agency will be created, which will report to the US Securities and Exchange Commission (SEC), similar to the US Financial Industry Regulatory Authority (FINRA). The sources mentioned that the plan is currently under review by White House Chief of Staff Susie Wells. This approach will provide more certainty to top AI labs. For example, Anthropic PBC was hit by US export controls last month, leading to the temporary suspension of their Fable 5 and Mythos 5 models; OpenAI also made significant adjustments before releasing their latest Sol model in response to government requests. Both companies have expressed objections to the government's actions, arguing that the intervention was excessive compared to the security issues identified by US officials. The sources mentioned that the proposal aims to appease two groups: Wall Street companies seeking to minimize AI network security risks, and Silicon Valley companies frustrated by what they see as inconsistency in government regulation of this emerging technology. They stated that the framework is still in the discussion stage and may change, allowing the two industries to play a greater role in setting security standards together. One source stated that Trump has not reviewed the plan yet, but officials have accelerated their work after the release of a new Chinese model triggered a sell-off of AI-related stocks, in order to provide more clarity to the AI industry. Wells typically only intervenes in policy debates (such as tariffs or AI policy) when there is political risk for the President and extensive coordination is needed between different officials and agencies. A White House official mentioned that the government is considering multiple proposals to strengthen the US's position in artificial intelligence and enhance cybersecurity. Discussion on AI regulation comes amid increasing competition with China for global leadership in this technology. On Friday, Chinese AI startup Moon's Dark Side launched its new Kimi K3 model, which can rival the more expensive systems from OpenAI and Anthropic, raising new questions among investors about whether the US industry will continue to heavily invest in AI chips and data centers. Venture capitalist and former White House AI czar David Sachs, who continues to advise Trump on technical matters, hinted that the new Chinese model is "worrisome" and is a result of the US "self-restraint," including pushing for government approval of AI models. Since the release of Anthropic's Mythos model (which has capabilities that have officials and business leaders on alert), the Trump administration has been pondering the level of intensity needed to address AI security risks. Last month, the US took a stricter approach, preventing foreign nationals from accessing Anthropic's top AI models. However, the latest proposal is more in line with an earlier executive order by Trump, outlining a voluntary review system created in partnership with AI companies. In general, the US plan aligns with the policy recommendations made earlier this week by Google DeepMind CEO Demis Hassabis. Hassabis is set to meet with policymakers in Washington next week to advocate for his plan. His proposal has garnered praise from Microsoft CEO Nadella, OpenAI CEO Sam Altman, and SpaceX CEO Musk, who are usually competitors. Hassabis's plan suggests that an independent committee of technical experts review AI models as part of a standard regulatory body within the federal government. He suggests industry funding for this self-regulatory organization to cover talent and computing costs, which will collaborate with government agencies and US national labs to develop testing protocols for areas such as cybersecurity and biothreats. In a post outlining his plan on X (formerly Twitter), Hassabis likened the regulatory body to FINRA, which is an independent, industry-funded regulatory body for brokerage firms, not technically part of the US government. Its authority comes from the SEC, which essentially outsources regulatory work to FINRA. It is currently unclear what type of model evaluations the government's proposal will require, how the new agency will be funded, and what role the SEC will play in overseeing AI companies. Both OpenAI and Anthropic are private companies, but both are considering the possibility of going public within the next year.