US Weighs Tightening Oversight of AI – NYT

(SeaPRwire) –   The White House is reportedly considering oversight of new artificial intelligence models prior to release, signaling a potential shift from its previously hands-off approach

The White House is exploring the idea of reviewing new AI models before their official launch, according to the New York Times, which cited sources on Monday. This marks a possible departure from its current noninterventionist policy.

As per the report, the administration of U.S. President Donald Trump is contemplating the formation of an AI working group that would include government officials and technology executives to examine oversight options, including the possibility of establishing a formal review process.

Sources indicate that this initiative aims to prevent political repercussions from a potential cyberattack enabled by AI, while also allowing officials early access to new models to evaluate whether they could provide cyber capabilities beneficial to the Pentagon.

The proposed system could grant the government early access to advanced AI models, although it would not necessarily have the authority to block them. Officials reportedly discussed these plans last week with representatives from Anthropic, Google, and OpenAI.

This approach may resemble the UK’s AI Security Institute, which assesses advanced models for risks and offers guidance to the government on implementing safeguards. In the U.S., major AI labs already voluntarily test models through the federal Center for AI Standards and Innovation, but the agency takes a more business-friendly stance, framing safety measures as a means to build public trust.

It remains uncertain how the reviews would operate or which agencies would oversee them. Any regulatory mechanism would likely be introduced via executive order, though no timeline has been established.

Implementing formal review requirements would represent a reversal of Trump’s AI policy. Since returning to office, he has advocated for minimal regulation, describing AI as “a beautiful newborn baby” essential for competing with China and cautioning against “politics or stupid rules” that could impede its development.

The New York Times reports that this policy shift gained urgency after Anthropic revealed a powerful new model, Claude Mythos, which it decided not to release publicly due to concerns it might expose critical software vulnerabilities.

The debate also follows a dispute between Anthropic and the Pentagon, triggered when the company refused to relax restrictions on surveillance and autonomous weapons. The Department of War designated it a “supply-chain risk,” excluding it from defense contracts—a decision Anthropic is now contesting in court.

The Pentagon has been rapidly expanding its use of AI in military operations, recently securing agreements with Google and OpenAI for classified AI systems.

Concerns about AI-related risks have been growing in the U.S. A Pew Research Center poll from last year found that 50% of Americans expressed greater concern than excitement about AI, up from 37% in 2021. A March Gallup poll showed Gen Z sentiment shifting negatively, with declining optimism and rising anger.

Public skepticism toward the Pentagon’s AI initiatives remains high. A February ITIF survey found that 79% of Americans believe a human should make final decisions regarding lethal force, while 75% stated AI is not reliable enough for life-or-death applications. The same poll indicated broad support for imposing limits on tech companies, with 67% agreeing that firms should restrict how their products are used—even by the government.