On June 2, 2026, President Donald Trump signed a long-awaited executive order on artificial intelligence and cybersecurity, marking a significant—if cautious—step in the U.S. government’s approach to the rapidly evolving world of advanced AI. The move comes amid mounting concerns over AI-driven cybersecurity risks, a spate of high-profile industry developments, and deep divisions within the administration over how much to regulate this powerful technology.
The order, which Trump signed privately after abruptly canceling a more stringent version just weeks earlier, directs federal agencies to shore up cyber defenses and asks AI companies to voluntarily submit their most advanced models for government assessment up to 30 days before public release. According to Axios, the order is a compromise: it shores up the country’s cyber defenses in the face of models like Anthropic’s Mythos, which recently exposed alarming software vulnerabilities, but stops short of imposing mandatory government licensing or preclearance for AI model development and release.
“Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies,” the executive order states. “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.”
Trump’s new directive arrives at a pivotal moment for the AI industry. Just one day before the order was signed, Anthropic confidentially filed with the Securities and Exchange Commission for an initial public offering, according to CNBC. Rival OpenAI is reportedly preparing its own IPO, and Elon Musk’s SpaceXAI is poised to debut on the public market as soon as next week, with a potential valuation that could exceed $1 trillion. The stakes, both economic and strategic, have never been higher.
At the heart of the order is a voluntary framework: AI developers are “encouraged” to provide the federal government and select critical infrastructure operators with up to 30 days of pre-release access to their most advanced models. This is a notable reduction from the 90-day window initially proposed in earlier drafts, reflecting industry pushback and a desire to avoid stifling innovation. According to Nextgov/FCW, the order’s language is clear: “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models.”
This hands-off approach is a marked shift from the Biden administration’s push for tighter federal oversight. Trump’s team, wary of overregulation, has repeatedly emphasized the need to keep American companies competitive, especially against rivals like China. Yet, the emergence of powerful AI models with sophisticated cyber capabilities—most notably Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber—has forced the administration to acknowledge the growing risks.
According to NPR, the Mythos Preview model, which Anthropic limited to a select group of companies as part of its Project Glasswing initiative, became a focal point for both government officials and Wall Street earlier this year. Its demonstrated ability to identify and exploit software vulnerabilities set off alarm bells in both Silicon Valley and Washington. The Department of Defense, in fact, labeled Anthropic a supply chain risk shortly before Mythos’s release, effectively banning its use in defense contracts—a designation that Anthropic is now challenging in court.
The executive order also includes a range of directives aimed at strengthening the government’s own cyber defenses. Within 30 days, national security agencies—such as the Department of Defense, the Treasury Department, the National Security Agency (NSA), the Cybersecurity and Infrastructure Security Agency (CISA), and the National Institute of Standards and Technology (NIST)—must develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models. This process will help determine when a model should be treated as a “covered frontier model,” a designation that triggers additional scrutiny and safeguards.
Federal agencies are also instructed to secure both Defense Department and civilian networks within 30 days. The order calls for the creation of an AI cybersecurity clearinghouse to review and share information on vulnerabilities, and it directs the Office of Management and Budget to identify existing federal grant funding that could support AI vulnerability-detection efforts within the same timeframe. Meanwhile, the Office of Personnel Management is tasked with boosting cyber hiring through the U.S. Tech Force, a program launched in December, which has thus far onboarded just 10 employees.
Behind the scenes, the order reflects intense internal debate and industry lobbying. Venture capitalist David Sacks, a longtime ally of Elon Musk and former White House AI czar, played a pivotal role in shaping the final language. Sacks and other tech leaders, including Musk and Meta CEO Mark Zuckerberg, reportedly lobbied the Trump administration last month to water down the original executive order. According to sources cited by Axios, Sacks was instrumental in securing the shorter 30-day pre-deployment testing window and ensuring that the order would apply only to advanced models, not the broader universe of AI systems. The abrupt cancellation of the earlier, stricter order was attributed in part to Sacks’s involvement, as well as Trump’s own concerns about hurting American competitiveness.
“The administration’s approach to AI has shifted in recent months,” Nextgov/FCW reports, “amid the emergence of Anthropic’s Mythos, a powerful cybersecurity-focused AI model that has become a major driver of government discussions, as officials grapple with how advanced AI systems can rapidly uncover vulnerabilities across computer networks.” OpenAI’s GPT-5.5-Cyber, which also demonstrated advanced cyber capabilities, has only heightened those concerns.
Yet, the order remains thin on specifics and relies heavily on voluntary cooperation from industry. While some see this as a pragmatic balance between innovation and national security, others worry it leaves the door open to significant risks. The administration has acknowledged that any binding regulations would have to come from Congress, not the executive branch.
The new order is also notable for what it does not do. Unlike some early proposals, it does not create any mandatory government licensing, preclearance, or permitting requirements for the development or release of new AI models. Instead, it aims to establish a “classified benchmarking process” and a voluntary coordination clearinghouse between the government, AI companies, and critical infrastructure operators. The Commerce Secretary, working through the Director of NIST, will assist in developing this process, while the NSA will play a key role in determining which AI systems meet the threshold for additional oversight.
As the U.S. government attempts to navigate the fine line between fostering innovation and protecting national security, the executive order signed this week represents a cautious, industry-friendly step. Whether it will be enough to address the mounting risks posed by advanced AI remains to be seen—but for now, the White House is betting on collaboration over compulsion.