Trump Administration's AI Self-Regulatory Executive Order Faces Obstacles
Nashnova编辑部
A draft executive order circulating inside the Trump administration would create a FINRA-style self-regulatory body for frontier AI companies, but strong opposition from former White House AI czar David Sacks has stalled the effort — exposing an unresolved split over how to govern AI.
What does the draft order propose?
The core idea: a FINRA-like body — FINRA is the self-regulatory organization that oversees U.S. broker-dealers — but for AI, tasked with pre-release testing of frontier models and reporting results to federal agencies.
This means → if enacted, companies like OpenAI and Google would face an industry-run checkpoint before shipping new models, with the government kept in the loop.
The concept traces to Google DeepMind CEO Demis Hassabis, who publicly called for a FINRA-model AI standards body last month. Treasury Secretary Scott Bessent had also floated a similar idea.
Why has it stalled?
Final sign-off requires multiple senior officials including the president; progress has visibly slowed.
The key obstacle: David Sacks, former White House AI and crypto czar. Though he left government earlier this year, he still advises the administration as a venture capitalist — and his opposition carries weight.
On his "All-In" podcast, Sacks called the FINRA-style approach a "terrible idea" and likened it on X to "a DMV for AI" — arguing rules will only pile up against theoretical risks.
What alternative does Sacks want?
He backs the MPA (Motion Picture Association) model — an industry-run rating system that operates independently and does not report to the government.
In plain terms = FINRA-style means "self-regulate + report to government"; MPA-style means "self-regulate, government stays out." The core divide: whether government gets a seat at the table at all.
This reflects Silicon Valley venture capital's instinctive resistance to any oversight framework with a government reporting line — even one branded as "self-regulatory."
Where does the Trump administration actually stand on AI regulation?
The overall lean is deregulatory: multiple Biden-era AI rules have been rolled back, and the U.S. AI Safety Institute was renamed and downsized.
But a counter-signal emerged: a June executive order established a voluntary framework for frontier-model evaluation, negotiated over months and still unpublished as of early August.
This means → the administration has not resolved the tension between "hands-off entirely" and "light-touch oversight." Hassabis's proposal overlaps heavily with the June framework, and how the two would coexist remains undefined.
What does this mean for the industry?
In the near term, frontier AI companies (OpenAI, Google, Anthropic) face no new pre-release review requirements — the order is on hold.
But as Anthropic and others ship increasingly capable models, some officials are pushing for stronger safety guardrails. Regulatory pressure has not disappeared.
In plain terms = the question is no longer "regulate or not" — it's an internal power struggle over "regulate how." The outcome hinges on whether the administration can reach consensus on a governance model.
市场有风险，内容仅供研究参考，不构成投资建议。