White House Weighs Review Process for Advanced AI Models Before Release

The Trump administration is considering government oversight for new artificial intelligence models, including a formal review process before public release, according to a New York Times report corroborated by Reuters. The discussions include a possible executive order that would create an AI working group of government officials and technology executives to examine oversight procedures, Reuters reported.

The Details

Reuters reported that the possible shift is under discussion but has not been announced as policy by the White House. The New York Times reported that White House officials discussed the plans last week with executives from Anthropic, Google and OpenAI.

The approach under consideration would create a formal channel for reviewing new AI systems before they reach the public, according to Reuters. The reported executive order would establish an AI working group that brings together government officials and technology executives to study oversight procedures.

Reuters reported that a White House official declined to confirm or deny the report. The official said, "Any policy announcement will come directly from the president. Discussion about potential executive orders is speculation."

The reported deliberations follow concerns about Anthropic's new model Mythos, Reuters said. Cybersecurity experts warned that Mythos could materially improve the ability to find and exploit software vulnerabilities, according to Reuters.

The Verge separately reported that some officials are interested in securing government first access to major new AI models even if that access does not block their eventual release. The Verge tied the internal concern to the risk of an AI-enabled cyberattack and cited unnamed sources describing worries about "political repercussions if a devastating A.I.-enabled cyberattack were to occur."

The reporting leaves unresolved whether the administration is moving toward mandatory pre-release review or weighing a narrower access arrangement. The New York Times and Reuters reported that a formal review process is being considered, while the White House official quoted by Reuters said discussion of potential executive orders remains speculative until any announcement comes from President Donald Trump.

Reuters described the possible oversight plan as a policy reversal from the administration's earlier posture toward AI. The reported talks would place the White House back in direct conversation with leading AI developers about release procedures for advanced models.

Context

Reuters noted that Trump took a comparatively hands-off stance on AI earlier in his term. Reuters reported that he revoked Joe Biden's 2023 AI executive order on his first day in office in 2025.

Reuters said Biden's revoked 2023 order had required developers of high-risk AI systems to share safety test results with the U.S. government before public release under the Defense Production Act. The new discussions reported by The New York Times and Reuters center on whether the Trump administration should again require some form of government review before advanced AI systems are released.

Reuters also reported that the White House unveiled an AI policy for Congress in March urging federal pre-emption of state AI rules while focusing on child safety and energy-cost protections. That March policy is part of the record Reuters cited while describing the administration's current discussions over federal oversight of AI models.

The companies named in the New York Times report are Anthropic, Google and OpenAI. According to Reuters, the specific concern now includes Anthropic's Mythos model and warnings from cybersecurity experts about its ability to improve vulnerability discovery and exploitation.

What's Next

No executive order has been announced, and the White House official quoted by Reuters said any policy announcement would come directly from President Donald Trump. Until then, the reported options remain under consideration rather than finalized policy.

The issues still not described in public reporting include the exact review criteria, the model thresholds that would trigger review, and whether any process would be mandatory for all frontier AI models. Reuters, The New York Times and The Verge all framed the discussions as active deliberations rather than a completed rule.

If the administration proceeds, the reported working group would bring government officials and technology executives into the same process for examining oversight procedures. Reuters reported that this working-group structure is part of the possible executive order under discussion.