The Trump AI EO strikes a compromise to balance innovation with accountability

COMMENTARY: The debate over how the federal government should regulate AI has emerged as one of the defining policy questions of the modern era. At the center of that debate: whether the White House should implement, strengthen, pause, or repeal executive orders (EOs) governing AI development and deployment.

On May 21, the White House paused plans to sign an EO on AI. Shortly thereafter, a revised version was released June 2 that asked AI companies to voluntarily let the government review frontier AI models at least 30 days before their public release. The administration wanted 90 days, the industry 14, thus the 30-day compromise.

[SC Media Perspectives columns are written by a trusted community of SC Media cybersecurity subject matter experts. Read more Perspectives here.]

The EO’s shifting trajectory highlights a fundamental tension between two competing priorities: accelerating innovation to maintain geopolitical dominance and establishing safeguards to protect society from the unintended consequences of increasingly powerful technology.

The case for deregulation: America First

Supporters in the industry of limiting federal oversight argue that the U.S. cannot afford to slow down in the global AI race, particularly as China and Russia aggressively pursue advanced AI capabilities. From this perspective, excessive regulation risks suppressing innovation, burdening startups with compliance costs, and weakening America’s strategic advantage in one of the most consequential technological revolutions in history.

Advocates of pausing or repealing AI EOs believe that a decentralized, market-driven approach lets companies innovate faster, attract investment, and develop transformative technologies before geopolitical adversaries gain the upper hand. They warn that overregulation could drive talent, capital, and research offshore to countries with fewer restrictions, ultimately undermining U.S. leadership in AI development.

These arguments carry legitimate weight. In an environment defined by rapid iteration and exponential technological advancement, policymakers face genuine pressure to avoid imposing constraints that could slow progress or hand competitive advantage to adversaries who operate without equivalent restrictions.

While there’s some real concern here, it’s also incomplete.

Proponents of maintaining and strengthening federal AI EOs favor a precautionary, standardized approach designed to mitigate existential and societal risks, protect civil liberties, and establish the U.S. government as a central authority in global AI governance.

Critically, this position does not argue against rapid AI development. It argues that how society manages the deployment and operational use of advanced systems once they move beyond research labs and into the public matters as much as the pace of development itself. The security community has understood this principle for decades. Building capability faster than we can defend it is not a competitive advantage: it’s a liability.

Advanced large language models in the hands of hostile nation states, cybercriminal organizations, or rogue proxy actors present significant national security concerns. Recent developments demonstrate that these concerns are no longer theoretical. China has demonstrated the operational use of Anthropic’s Claude model to simultaneously target more than 20 victims in coordinated cyber operations. Using agentic AI capabilities, the system reportedly reasoned in real time about attack vectors, adapted dynamically to defensive countermeasures, and modified its behavior to evade detection mechanisms.

The implications of this incident extend beyond the specific campaign. It demonstrates that adversaries already view advanced AI systems as offensive capability multipliers capable of dramatically increasing the scale, speed, sophistication, and persistence of cyberattacks.

Two philosophies, one threat landscape

The regulatory debate reflects genuinely different governing philosophies. But adversaries do not pause their operations while Washington deliberates.

The signing of this most recent AI EO establishes a minimum viable safeguard.

Pausing or repealing these measures removes critical layers of visibility and coordination between the private sector and the government agencies responsible for national security and public safety. Without standardized oversight, companies may prioritize speed to market over long-term security considerations, leaving the public vulnerable to unchecked and potentially dangerous AI capabilities.

The question is not whether the U.S. should lead in AI. It should, and it must. AI leadership will shape the future of economic power, national defense, global influence, and technological dominance for decades to come. The real question for us: whether we will build America’s leadership on a foundation strong enough to withstand exploitation, or whether the widening gap between technological capability and meaningful governance will become the vulnerability our adversaries are waiting to exploit.

In an era defined by exponential technological acceleration, pausing the development of guardrails or basic oversight will not make society safer. It will simply expand the attack surface while enabling threat actors to increase their level of exploitation.

Policymakers must continue to ensure that innovation and security evolve together. We should not view responsible governance as an obstacle to innovation, but as the framework that allows innovation to scale safely, sustainably, and strategically. Without that balance, the U.S. risks creating world-changing technologies that are insufficiently protected from those determined to use them against us.

James Turgal, vice president, global cyber risk and board relations, Optiv

SC Media Perspectives columns are written by a trusted community of SC Media cybersecurity subject matter experts. Each contribution has a goal of bringing a unique voice to important cybersecurity topics. Content strives to be of the highest quality, objective and non-commercial.