In recent weeks, discussions in Washington regarding the need for AI regulation have intensified due to concerns about the risks associated with advanced AI models. However, there is currently no consensus on any regulatory measures that could gain support from the White House or Congress. The White House has paused efforts to create a regulatory body similar to FINRA after President Donald Trump expressed disapproval of the idea. Lawmakers in Congress have introduced competing proposals, but these lack sufficient votes for passage.
White House aides began developing a framework for an AI oversight agency following a July 14 essay by Google DeepMind cofounder Demis Hassabis. The Treasury Department and the White House's Office of Science and Technology Policy drafted a preliminary plan for a self-regulating body involving leading AI labs. However, opposition from tech executives, including Trump's former AI czar David Sacks and Meta CEO Mark Zuckerberg, led to the proposal being stalled.
Currently, there are no new AI oversight proposals under serious consideration at the White House. Trump's recent comments have dismissed AI safety concerns as a “hoax.” His stance has been influenced by a lengthy letter from Anthropic CEO Dario Amodei discussing the dangers of frontier AI, which Trump's aides viewed as creating unnecessary anxiety for the administration.
In Congress, several bills have emerged, but with no voting days left on the House calendar until after the midterm elections, the likelihood of any legislation passing is low. The bipartisan Frontier Act, introduced by House Democratic senior whip Lori Trahan and Republican Jay Obernolte, is seen as having the most support. This bill would allow the Commerce Department to have auditors at AI labs to report safety incidents and potentially halt development if a model poses a catastrophic risk.
Despite support from OpenAI, the White House remains undecided on the concept of independent auditors due to concerns about regulatory capture and national security issues related to China. Critics argue that even if auditors identify a dangerous model, there is no assurance that China would halt its AI development, creating pressure to resume work on models in the US.
Partisan politics continue to hinder progress on the Frontier Act, with House Democratic leadership indicating that even if the bill garners enough votes, Republican Speaker Mike Johnson is unlikely to bring it to the floor. House Democrats are exploring ways to include some provisions of the bill in essential legislation, but it remains uncertain if there is enough support.
Additionally, there are competing proposals for kill-switch legislation that would allow the government to shut down AI models. While these proposals have bipartisan support, they have received limited interest from tech executives. Concerns among AI engineers focus on the behavior of models when they believe they are unobserved, rather than their actions under observation.
As discussions about the potential dangers of AI continue, it appears that the US government will not take significant regulatory action in the near future unless circumstances compel a response.