How US federal AI policy has gone from implausibly libertarian to increasingly draconian and opaque, and how to fix it, including using independent auditors
35 thoughts on what has happened and what America should do — Nothing below is an official or unofficial view of OpenAI.
HyperdimensionalDean W. Ball
Context & Ripple Effects
Related coverage shows OpenAI pressing for a workable national AI-policy framework while warning that a large volume of state bills could fragment the US market. More recently, its policy leadership has pursued a “reverse federalism” strategy: encouraging blue states to set safety rules while Washington remains stalled.
This intervention lands against that split backdrop. Its call for independent auditing is a concrete alternative to both minimal federal constraints and an opaque, highly restrictive federal approach.
First-order effects
The proposal puts independent auditors at the center of the federal AI-policy debate, offering a distinct oversight mechanism for AI-system governance.
OpenAI and other AI developers face a clearer policy choice in public debate: accept externally verifiable oversight or continue navigating a patchwork of state-led safety initiatives.
Second-order effects
State policymakers pursuing AI safety rules may treat independent-audit requirements as a reusable baseline, potentially making them more compatible across jurisdictions.
An audit-centered approach would elevate demand for credible assurance processes and could shift competitive pressure toward developers able to document and submit systems to outside review.
Third-order effects
If federal policy remains unsettled, the US may continue to develop AI governance through a mix of state standards and company-backed proposals rather than one settled national regime.
The durable fault line is likely to be not whether AI is governed, but whether governance relies on transparent, independently testable controls or discretionary and less visible federal restrictions.
The trend: US AI governance is moving from broad arguments over deregulation versus restriction toward a contest over who sets enforceable safety standards and how independently they can be verified.
It seems both foolish and wrong to frame the current chaos as the fault of *safety* activists, who've been arguing for clear, defined rules for years now. It's much more the fault of people like Andreessen: their completely deregulatory stance means those clear rules don't [image…
In a matter of weeks, U.S. federal AI policy has gone from implausibly libertarian to increasingly draconian and opaque. Today, over 35 distinct observations, I analyze how we got here and offer the most succinct statement I can about what exactly I propose we should do next. [im…
While I have critiques of the export controls and other recent actions, there is a core truth that shouldn't be overlooked: before the last few weeks, it was fair to question whether the United States took catastrophic AI risk seriously. Today, the answer is unambiguous: “yes.”
According to so many experts here, the Federal Government is blameless for its problematic policies, and a 5,000-person company in fact deserves all that blame. Interesting!
This is what's causing Anthropic to aggressively beg for govt protection (see below). Customers are finding cheaper alternatives. Keeping employees requires continuing ultra-rich secondaries ($$$) that are dependent on revenue growth. When you can't win on the field go to DC.
The Trump administration has had a major pendulum swing on regulating AI from anything goes to saying they have to approve any user of GPT 5.6 and non-US citizens can't get Claude Fable. — This is now a major problem for AI labs after spending billions to train their models.
It is bad for AI *deployment,* which is different from development. On a longer timescale (months, not weeks), yes, that will hurt funding + slow compute purchases etc. + slow development. TBD if it gets sorted before then. For now, rest assured the GPUs are humming
This is not an AI pause. It's the government thrashing around after realizing that the people who told them AI was NBD were wrong, and being severely understaffed. AI development is not even close to stopped RN + this is not what you'd do if you wanted it to stop
House Republican leadership and the trump Administration made a decision to let the AI industry run wild. That has resulted in massive public backlash. Now the trump folks realize, oh shit, the American people were right. We need reasonable AI guardrails. November is coming.
There's a lot of sloppy thinking around open models. You can ban them and make it impossible for US companies to use them, but this won't stop A) global open model progress B) bad actors using them So what exactly is gained by banning open models, including those from China?
Getting a highly unsurprising set of people RTing this with versions of 'if you people hadn't warned of an underlying problem and called for laws to be passed and preparations to be made, which we ensured didn't happen, we wouldn't have this ad hoc thing, this is your fault.'
Our new AI policy is that the White House decides ad hoc, for whatever reasons it likes, who does and does not get access to frontier intelligence. This seems rather maximally terrible.
If you are on the verge of AGI or ASI, why isn't your model smart enough to recognize espionage distillation in real time? You say “cure cancer in a few years.” Isn't sniffing illicit distillation quite a bit easier than curing cancer? Why write letters to DC? Just use AGI.
It's been two weeks and the leading US model, Fable, is still banned by the White House with no explicit Congressional Authorization, no written rule that was broken, and no technical justification. In the meantime, our Chinese adversaries released a model that rivals our best.
@Miles_Brundage No one said that AI was no big deal, it has nothing to do with them being understaffed, and the question isn't whether AI development is stopped or not, that was never going to happen no matter what Doomers dreamed of, the question is whether the discourse from Do…