OpenAI Backs Trump Order Requiring 30-Day Federal AI Model Reviews
Updated
Updated · CNBC · Jun 5
OpenAI Backs Trump Order Requiring 30-Day Federal AI Model Reviews
3 articles · Updated · CNBC · Jun 5
Summary
OpenAI said it will join Donald Trump’s voluntary executive-order framework, giving the federal government access to assess AI models 30 days before release.
The order, signed Tuesday, calls for benchmarking advanced cyber capabilities to determine when a system qualifies as a “covered frontier model.”
George Osborne, OpenAI’s head of countries, said the company had already pushed for governments to track AI safety and security risks rather than waiting to be compelled.
At SXSW London, Osborne argued governments need strong but flexible regulators, underscoring a broader debate over how to oversee increasingly powerful frontier AI systems.
Will a 30-day government check on new AI models actually improve security or just slow down American innovation?
Can a secret government benchmark truly keep pace with AI models whose capabilities are constantly and unpredictably evolving?
As the U.S. begins classified AI reviews, what will prevent this from sparking a new global AI arms race?
Trump’s 2026 Executive Order on AI: Voluntary Federal Review, Critical Infrastructure, and Industry Response
Overview
Amid rising concerns about the security risks of advanced AI, highlighted by Anthropic's decision to limit its Mythos Preview model after it showed the ability to find and exploit software vulnerabilities, President Donald Trump signed an executive order on June 2, 2026. This order created a new voluntary federal review framework for advanced AI models. The framework allows government agencies early access to powerful AI systems before public release, using a new benchmarking process. Key agencies like the NSA and DoD help decide which models need scrutiny, aiming to identify and address security issues before these technologies reach the wider market.