Trump issues executive order to evaluate AI models prior to their launch.
President Donald Trump has enacted an executive order aimed at establishing a “voluntary framework” for AI companies to share their advanced models with the federal government prior to their public release. This initiative is designed to foster secure innovation and enhance the cybersecurity of essential infrastructure.
The order acknowledges the success of the US AI industry, attributing it to a lack of overly restrictive regulations, while also recognizing that new AI advancements introduce security challenges. It instructs various federal agencies to develop a framework for evaluating the advanced cyber capabilities of AI models before they hit the market. Companies will have the option to share their models with the government ahead of release, with the potential for confidentiality protections if they choose to participate. Additionally, the federal government is tasked with bolstering cyber defenses for AI, particularly concerning critical infrastructure.
This order follows a last-minute decision by Trump to delay signing a previously planned executive order, which he believed could hinder competition with China. The earlier draft would have permitted AI companies to voluntarily share their models 14 to 90 days before release, while the current version shortens that window to up to 30 days. Major players like Google, Microsoft, and xAI have already agreed to allow pre-release reviews by the Commerce Department’s Center for AI Standards and Innovation (CAISI).
OpenAI and Anthropic had previously committed to sharing their models with CAISI back in 2024 under President Joe Biden’s initiative for AI safety measures. However, the Trump administration had previously downplayed safety concerns and adopted a more hands-off stance under former White House AI advisor David Sacks. The newly signed order clarifies that it should not be interpreted as mandatory licensing or preclearance, yet it indicates a shift towards increased oversight of AI companies.
One possible reason for this change could be Anthropic’s limited rollout of its powerful Mythos model in April, which reportedly identified “thousands of high-severity vulnerabilities” across major operating systems and web browsers. This development may have opened the door for improved relations between Anthropic and the administration, especially following a legal dispute with the Pentagon regarding the use of AI in autonomous weapons and mass surveillance.
The order has received positive feedback, even from groups that typically oppose restrictions on state AI regulations. Brad Carson, President of Americans for Responsible Innovation, remarked that the White House is now taking AI vulnerabilities seriously. Brendan Steinhauser, CEO of Alliance for Secure AI, expressed satisfaction with the administration’s acknowledgment of the risks associated with these models and both he and Carson urged Congress to establish mandatory protections.
