The White House is preparing to drop a new executive order that would reshape how AI companies bring powerful models to market. The core requirement: developers would need to share their AI models with the federal government a full 90 days before releasing them to the public.
The proposed executive order targets the intersection of AI development and national security, specifically cybersecurity risks posed by increasingly capable AI models. The concern isn’t theoretical. As AI systems grow more powerful, so does their potential for misuse in offensive cyber operations, from automating vulnerability discovery to generating sophisticated phishing campaigns at scale.
The framework is being described as voluntary rather than mandatory. The 90-day pre-release sharing requirement is the headline provision, but it sits within a broader strategy. The goal is to give federal agencies enough lead time to assess whether new AI models introduce national security vulnerabilities before they’re deployed in the wild.
The order is reportedly still under discussion, meaning the final version could look different from what’s been circulated.
A pattern of AI executive orders
