AI
Trump's rejection of AI oversight stalls all regulatory efforts in Washington
Despite mounting alarm over frontier AI risks, the White House has abandoned its agency proposal and Congress lacks the votes or political will to pass meaningful safeguards.
WIRED — AI
Lawmakers and regulators have been discussing the need for stronger guardrails around advanced AI systems, but recent developments suggest that meaningful federal action remains far off. The White House has quietly shelved plans to create an independent oversight body, while competing congressional proposals face steep political obstacles.
The Failed White House Initiative#
Earlier this year, officials at the Treasury Department and the White House Office of Science and Technology Policy drafted preliminary plans for an AI regulatory agency modeled on FINRA, inspired partly by Demis Hassabis of Google DeepMind. The framework would have allowed AI developers to self-regulate frontier models while maintaining government oversight.
That effort collapsed after tech leaders—including Meta's Mark Zuckerberg and Trump's former AI advisor David Sacks—persuaded President Trump to oppose it in August. The plan has since languished without new proposals under serious consideration at the administration level. Trump has since hardened his stance further, calling AI safety concerns a "hoax" and rejecting any form of explicit regulation.
Congressional Proposals Stuck in Gridlock#
With the White House inactive, lawmakers have introduced several competing bills, but none appears positioned for passage. The most prominent measure—the bipartisan Frontier Act, backed by representatives from both parties and supported by OpenAI—would authorize the Commerce Department to embed auditors within AI laboratories. These auditors could report safety incidents and halt development of models deemed catastrophic risks.
Despite this backing, the proposal faces resistance from Trump's advisors who worry that pausing dangerous AI development unilaterally could leave the U.S. vulnerable if China does not follow suit. Speaker Mike Johnson has shown no interest in bringing the bill to the floor, and even within the Democratic caucus, support remains incomplete.
Alternative measures include kill-switch legislation that would permit government-mandated shutdowns of AI systems. Both Democratic and Republican versions exist, yet tech executives have expressed little enthusiasm for these approaches, questioning whether turning off a model after it has already caused damage would accomplish anything meaningful.
