Trump administration plans AI executive order expansion
Trump postpones signing AI cybersecurity executive order hours before ceremony after pressure from tech billionaires and advisers opposing pre-release model testing frameworks.
Objective Facts
On May 21, 2026, Trump postponed signing an executive order on AI cybersecurity and model review, saying hours before the signing that he "didn't like certain aspects" of the order. The draft order would have created a federal process to evaluate frontier AI models before public release to identify cybersecurity risks, with a voluntary process for major AI companies to share models with the government 14 to 90 days before release. David Sacks, the former White House AI and crypto czar, successfully mounted a last-ditch lobbying effort to derail the order by calling Trump, with similar efforts from Elon Musk and Mark Zuckerberg. The order's development was prompted by Anthropic's April 2026 revelation of its Mythos model, which demonstrated extraordinary capability at finding network vulnerabilities and could pose major cybersecurity threats, with Anthropic limiting access to a handful of technology and Wall Street companies. Public opinion favored Trump's testing plan, with 47 percent of Republican voters strongly supporting it and over 70 percent believing in legally mandatory independent testing.
Left-Leaning Perspective
Prominent left-leaning voices who had been pushing for AI regulation expressed disappointment at Trump's postponement. According to Axios, those who had been pushing for AI regulation in Washington were initially relieved that the White House was finally going to make moves on AI and cybersecurity safety, but now face uncertainty about whether that will happen. Senator Gary Peters (D-Mich.), ranking member on the Senate Homeland Security Committee, specifically told Axios: "Starving CISA of personnel, resources and leadership in this high-stakes environment puts our homeland security and national defense at risk." Progressive Democrats framed the issue not just as cybersecurity but as part of a broader inequality problem. Congressional Progressive Caucus Chair Greg Casar stated: "Americans are rightly concerned that AI may make a few billionaires richer but lead to millions losing their jobs." Axios noted that five influential progressives—Senators Bernie Sanders and Elizabeth Warren, Representatives Alexandria Ocasio-Cortez and Ro Khanna—are taking the lead in shaping a more confrontational message on AI. Warren is proposing to tax AI companies and data centers, arguing in a TIME op-ed that "The American people deserve to share in the success of this technology." What left-leaning coverage emphasizes but right-wing outlets downplay is the national security dimension—that advanced AI models pose concrete, measurable cybersecurity threats that require coordinated federal oversight. A bipartisan letter to Trump marked an escalation in pressure on the administration, with lawmakers noting that Mythos identified "thousands of high-severity zero-day vulnerabilities" across major operating systems and web browsers.
Right-Leaning Perspective
Conservative and pro-deregulation voices celebrated Trump's postponement as protecting U.S. AI dominance and preventing government overreach. According to Axios, Trump, AI adviser David Sacks, and some technology executives discussed the order before the signing was delayed, with "The main reason the signing was delayed" being that Trump "just hates regulation," adding that Sacks also "hated it." David Sacks, described as a prominent AI "accelerationist" who believes federal regulation will harm U.S. innovation, successfully mounted a last-ditch lobbying effort to derail the order. The accelerationist framing holds that any government review process, even a voluntary one, could slow AI development and hand competitive advantage to China. One source told Axios that the order was "unnecessary" and "just something doomers wanted." According to Politico, Sacks told Trump that he feared the voluntary vetting would act as a de facto licensing regime, slowing down AI companies' releases of new models, and that a future administration might easily turn the voluntary procedure into a mandatory one. Crypto Briefing noted that whatever Google DeepMind, Microsoft, and xAI agree to voluntarily will likely become the de facto standard, with the companies writing the rules they already comply with, making formal regulation unnecessary. What right-wing coverage tends to omit is the national security case—that Mythos represents a new class of dual-use AI capabilities that even the companies themselves deemed too dangerous for unrestricted release. The Trump administration is continuing an ad hoc licensing process for Anthropic's Mythos model, which has been shared with the U.S. government and, under Project Glasswing, with a select handful of U.S. technology companies and financial institutions.
Deep Dive
The Trump administration's abrupt postponement of a voluntary AI safety review executive order on May 21, 2026, exposes a fundamental tension within Republican leadership between national security and deregulatory ideology. For context: Ever since Anthropic announced Mythos in April 2026, the White House has been scrambling to understand its hacking capabilities and what the administration's role should be in managing national security concerns, with the White House also eying executive actions to allow federal agencies to sidestep the ban on using Anthropic products. In a "post-Mythos world," the Trump administration appears to be reevaluating its hard line against AI safety and security measures it once shrugged off. The order itself was carefully designed to be non-threatening to industry—it was explicitly voluntary, requested only 90 days of pre-release access, and mirrored testing processes already occurring at NIST. Tech companies have been broadly supportive of AI model testing and guardrails, and leading frontier models already do voluntary testing through NIST's Center for AI Standards and Innovation. Yet even this minimal framework was unacceptable to an influential cluster of venture capitalists and AI executives. According to Axios, the main reason the signing was delayed was that Trump "just hates regulation," with Sacks also opposing it as "unnecessary" and "just something doomers wanted." What the right characterizes as "accelerationism" (prioritizing speed to market) conflicts directly with what the left sees as basic national security stewardship during a period of rapidly advancing dual-use capabilities. Crucially, public opinion cuts against both Trump's decision and his rhetorical framing: 47 percent of Republican voters strongly support pre-release testing, over 70 percent believe in legally mandatory independent testing, and 80 percent of Americans overall support such oversight. Yet Trump deferred to billionaire tech leaders rather than his own political base or the broader national security community. AI regulation splits Trump's base—he entered office supported by "move fast and break things" Silicon Valley billionaires like Sacks and Musk, but parts of his populist base favor stronger guardrails. The unresolved question going forward: will future AI incidents force Trump's hand again, as occurred in May, or will the accelerationist wing maintain veto power over any security-focused initiatives?