Summary
- President Donald Trump has enacted an executive order focused on enhancing U.S. cybersecurity through advanced AI technologies.
- The directive creates a voluntary framework for assessing and reviewing powerful AI models prior to their wider deployment.
- This decision follows a postponement of a similar initiative, where Trump expressed concerns that it might undermine the U.S. advantage in AI against China.
On Tuesday, President Donald Trump signed an executive order that aims to bolster cybersecurity in the United States by leveraging advanced artificial intelligence, while also fostering collaboration between federal agencies and prominent AI firms.
The executive order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," instructs federal agencies to expedite the adoption of AI-driven cybersecurity solutions, establish an AI cybersecurity clearinghouse, and implement a system for identifying advanced AI models.
“Advanced AI capabilities make our nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies), and components,” states the executive order. “As these capabilities evolve, my administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country.”
The order mandates that a classified review process be set up, wherein the National Security Agency will determine if advanced AI systems fall under the category of covered frontier models.
Developers will have the option to voluntarily submit these models to the government for evaluation “for a period of up to 30 days before they plan to release such models to other trusted partners.”
In May, Trump postponed the signing of a similar executive order, citing concerns that some aspects of the proposal could hinder U.S. AI advancement and diminish America's competitive edge over China.
Critics argue that the framework relies excessively on the voluntary compliance of the AI companies it intends to regulate. J.B. Branch, AI governance and technology policy counsel at the consumer advocacy nonprofit Public Citizen, remarked, “Models powerful enough to threaten cybersecurity and national security warrant real oversight. Congress and the administration should enact comprehensive federal AI legislation with enforceable safeguards, transparency requirements, independent testing, and meaningful protections for workers, consumers, children, and civil rights.”
The initiative to create an AI-centric executive order gained traction amid concerns regarding Anthropic's Claude Mythos model, which showcased an ability to identify software vulnerabilities and raised alarms among officials about the national security risks posed by increasingly sophisticated AI models.
In April, after the unveiling of Mythos, U.S. Treasury Secretary Scott Bessent and then Federal Reserve Chair Jerome Powell reportedly held a meeting with Wall Street bank executives, cautioning them about cybersecurity threats associated with the new AI model.
Despite these issues, Anthropic has proceeded to expand access to Mythos. On Tuesday, the developer announced it is broadening access to its Claude Mythos AI model through Project Glasswing, aimed at enabling tech and security firms, as well as governments, to identify and mitigate potential vulnerabilities before the model is publicly released—which the company indicated would occur “in the coming weeks.”
While the executive order is extensive, it also seeks to reassure AI developers that the new framework will not impose a formal approval process for the launch of new models. Additionally, it coincides with Trump's efforts to establish a federal regulatory framework for AI as numerous states advance their own legislation.
The order also emphasizes stricter enforcement against the criminal use of AI, which includes unauthorized access to any public or private information technology systems or “using AI agents to illegally obtain data or information that is subsequently utilized for unlawful purposes.”
Last month, federal prosecutors charged two individuals for employing AI to create and disseminate explicit images of women without their consent, marking one of the first significant enforcement actions under the new Take It Down Act.