By Maggie Eastland and Nancy Cook
The Trump administration is considering plans for an independent regulator to vet the safety of artificial intelligence models with industry input, after Silicon Valley leaders complained about the ad-hoc nature of recent US moves to slow the release of cutting-edge AI systems.
Treasury Secretary Scott Bessent helped develop the proposal, which would create an independent regulatory agency for AI that would report to the Securities and Exchange Commission, similar to the Financial Industry Regulatory Authority, according to people familiar with the matter. The plan is now being reviewed by White House Chief of Staff Susie Wiles, said the people, who spoke on condition of anonymity because it has not been made public.
Such an approach would offer more certainty for leading AI labs like Anthropic PBC, which last month was hit with US export controls that led it to temporarily disable its Fable 5 and Mythos 5 models, and OpenAI, which made significant changes at the government’s request before releasing its latest Sol model. Both companies objected to the government’s moves, calling them excessive relative to the safety issues identified by US officials.
The proposal aims to appease both Wall Street firms seeking to minimize AI’s cybersecurity risks and Silicon Valley companies frustrated by what they see as an inconsistent administration approach in overseeing the nascent technology, said the people. The framework, which remains under deliberation and is subject to change, would allow both industries to play a larger role in jointly setting safety standards, they said.
President Donald Trump has not yet reviewed the plan, one of the people said, however officials have accelerated work to give the AI industry more clarity following the release of a new Chinese model that’s fueled a selloff in AI-related stocks. Wiles usually only becomes involved in policy debates, like discussions over tariffs or AI policy, when they show signs of political risk for the president and require lots of coordination across different officials and agencies.
A White House official said the administration is considering multiple proposals to cement the US lead in AI and strengthen cybersecurity. The Treasury Department didn’t immediately respond to a request for comment.
The deliberations about AI oversight are unfolding amid intensifying competition with China for global leadership in the technology. On Friday, Chinese AI startup Moonshot unveiled its new Kimi K3 model, which rivals more expensive systems from OpenAI and Anthropic, sparking new questions among investors about whether industry in the US will sustain its spending spree on AI chips and data centers.
David Sacks, the venture capitalist and former White House AI czar who continues to advise Trump on technology, suggested the new Chinese model was “concerning” and the result of America “tying itself in knots,” including by pushing for government approval of AI models.
Since the release of Anthropic’s Mythos model, which features cyber capabilities that alarmed officials and business leaders, the Trump administration has grappled with how firm a hand it should take in addressing AI safety risks. Last month, the US took a more stringent approach in blocking foreign nationals from accessing Anthropic’s best AI models. The latest proposal, however, is more consistent with an earlier Trump order that outlines a lighter-touch, voluntary review system created in collaboration with AI companies.
Broadly, the US plan aligns with policy suggestions published earlier this week by Google DeepMind chief executive officer Demis Hassabis, who will be meeting with policymakers in Washington next week to lobby for his plan. His recommendations won praise from Microsoft Corp. Chief Executive Officer Satya Nadella, as well as OpenAI CEO Sam Altman and SpaceX CEO Elon Musk, who are typically bitter rivals.
The plan devised by Hassabis suggests a board of independent technical experts review AI models as part of a standards body overseen by the federal government. To pay for the talent and compute, he recommends that industry fund the self-regulatory organization, which would collaborate with government agencies and US national labs to develop model test protocols for areas like cybersecurity and biological threats.
In a post on X outlining his plan, Hassabis likened the watchdog to Finra, an independent, industry-funded regulator for brokerage firms that’s not technically part of the US government. It derives its authority from the Securities and Exchange Commission, which essentially farms out oversight to Finra.
It remains unclear what kinds of model assessments the administration’s proposal would call for, as well as how a new agency would be funded and what role the SEC would play in overseeing AI companies. Both OpenAI and Anthropic are privately held, though both companies are weighing initial public offerings within the next year.
