White House Weighs a Big Shift on AI Models

Washington's thinking on artificial intelligence appears to be changing quickly. After years of championing a mostly hands-off strategy, the Trump administration is expected to require that new AI models undergo a federal safety review before they hit the market, reports the New York Times. The outlet sees the shift as a "stark reversal." Axios reports the planning is far along, adding that the Pentagon would lead the safety evaluations. The rethink follows public anxiety over AI's impact and Anthropic's decision not to publicly release its Mythos system, which it warned could trigger a cybersecurity "reckoning."

An executive order under discussion would set up an AI working group of tech leaders and government officials to explore oversight options, including a safety vetting system similar to one Britain is building. Some in the administration want government first looks at cutting-edge models without blocking their release, both to avert an AI-enabled cyberattack and to assess potential military uses. The administration briefed leaders from Anthropic, Google, and OpenAI last week. As recently as July, President Trump spoke of the AI industry and declared, "We can't stop it with foolish rules."