Faris Adani
In a striking escalation of the ongoing technological rivalry between global superpowers, a former Obama-era national security official has put forth a provocative and potentially catastrophic proposal: to consider military strikes against Chinese data centers. This drastic measure, outlined in a new report, is suggested as a means to prevent China from achieving Artificial General Intelligence (AGI) before the United States, highlighting the deepening anxieties within Washington over the strategic implications of advanced AI development.
Jacob Stokes, a prominent figure previously serving as national security advisor to then-Vice President Joe Biden and at the Department of Homeland Security, articulated these recommendations in his report titled “Superpowers and AGI.” Published by the Center for a New American Security (CNAS), a highly influential Washington D.C. think tank specializing in national security policy, Stokes’ analysis calls for the United States to commence readiness drills in response to China’s accelerating AI advancements. These drills, he suggests, should encompass a spectrum of responses, extending even to direct military action against the critical infrastructure underpinning China’s AI ambitions: its data centers, as initially reported by the South China Morning Post.
At the heart of Stokes’ report lies the concept of AGI, or artificial general intelligence. This theoretical construct refers to a point where an AI model would not only match but potentially surpass human levels of reasoning, learning, and problem-solving across a broad range of tasks. While the scientific community remains divided on whether such a system is truly possible or imminent, Stokes’ report operates from the premise that AGI is an inevitable breakthrough. Consequently, if the US cannot be the first to achieve it, he argues, its military apparatus must prepare for extreme measures to prevent a rival from gaining an insurmountable strategic advantage.
Stokes, currently serving as the deputy director of the Indo-Pacific Security Program at CNAS, frames a scenario where the US becomes a “laggard state” trailing behind China’s AGI achievement. In such a predicament, he posits, the US could “try to sabotage the leading state’s AGI systems, either through physical infiltration or offensive cyber operations; that is, cyberattack.” This perspective underscores the deep-seated belief within certain policy circles that AI leadership is synonymous with global power, turning a technological race into a potential zero-sum game with profound geopolitical consequences.
Elaborating on the less aggressive, yet still highly provocative, options, Stokes considers cyberattacks as a preliminary step. “Cyberattacks to sabotage AGI might not provoke large-scale retaliation,” he suggests, based on the rationale that “cyberattacks are already to some extent considered normal state behavior below [the] threshold of war.” This argument positions cyber warfare as a potentially ‘deniable’ or ‘less escalatory’ form of aggression, allowing for disruption without immediately triggering a full-scale conventional conflict. However, the line between ‘below the threshold of war’ and an act of war itself is increasingly blurred in the digital age, making such calculations inherently risky, particularly when targeting critical national infrastructure like data centers.
Yet, an unprecedented cyberwar is merely one scenario Stokes outlines. He correctly observes that AI systems, particularly those capable of AGI, do not solely exist in the digital realm; they require immense physical computing power, relying on vast arrays of specialized hardware like Graphics Processing Units (GPUs) housed in physical data centers. Therefore, Stokes presents the most extreme and dangerous option: “kinetic attacks, meaning destroying things with missiles and bombs.”
“Data centers can be bombed with conventional munitions,” Stokes starkly writes. This chilling suggestion is not entirely novel in the discourse surrounding AI safety and national security. In 2023, prominent AI critic and alignment researcher Eliezer Yudkowsky similarly suggested that nations should be prepared to “destroy a rogue data center by airstrike” if it were developing potentially uncontrollable AI. This convergence of ideas from different corners of the AI discussion—from national security strategists to AI safety advocates—underscores the profound anxieties some hold regarding unchecked AI development, regardless of the target nation.
Stokes is not oblivious to the catastrophic implications of such an act. “Bombing would cross a major threshold that heretofore has not been crossed in US-China relations,” he acknowledges. He draws a parallel to the dawn of the nuclear age, stating, “Trying to think through the particulars of that will be especially important, in part because it will help policymakers… start to work backwards based on the unique nature of the technology, in the same way that in a past era, policymakers would learn about nuclear weapons.” This analogy is intended to convey the existential weight of AGI, suggesting it could be as transformative and dangerous as nuclear weapons, thereby justifying extreme preventative measures.
However, the comparison between AGI and nuclear weapons, while dramatic, reveals a critical flaw in Stokes’ premise. Nuclear weapons are tangible, empirically observable physical objects with well-understood destructive capabilities. AGI, in stark contrast, remains largely a theoretical construct, a “fantasy” for many leading researchers, with no clear consensus on its timeline, feasibility, or even definition. Basing potentially world-altering military strategies on such speculative technology carries immense risk. Critics quickly draw parallels to the “weapons of mass destruction” narrative that controversially justified the brutal invasion and occupation of Iraq in 2003, where the supposed threat was later revealed to be unfounded. The danger lies in preemptive strikes based on unproven threats, potentially leading to catastrophic real-world consequences.
Furthermore, Stokes’ report appears to overlook the nuanced reality of the US-China AI landscape. China’s approach to AI development is dramatically different from that of the United States, which currently leads the world in most large language model (LLM) benchmarks and foundational AI research. While China excels in AI application and integration into various industries and surveillance systems, the cutting edge of AGI research, if it exists, is arguably more concentrated in American labs. If anyone is actively “chasing AGI,” it is arguably the US. This raises a crucial question: how would the United States react if Chinese leaders were organizing war games around the prospect of bombing American data centers to halt its own leading AI development? The inherent hypocrisy of such a stance undermines the legitimacy of Stokes’ proposals.
Despite these critical counterpoints, Stokes remains undeterred in portraying China as the ultimate technological adversary. He warns of a “surprise technological breakthrough” by Chinese AI companies that could bring them closer to AGI, potentially enabling the production of “offensive cyber tools — in other words, cyber weapons.” Stokes, whose career has often focused on the economic strength of the US relative to its competitors, now extends this concern to the realm of advanced AI. His latest comments, however, mark a severe escalation, even within the hawkish circles of the US national security establishment, pushing the boundaries of what is considered a ‘defensive’ posture in the tech cold war.
Perhaps the most glaring omission in Stokes’ report is its blind eye to the present-day use of AI by his own country’s military. While fretting over speculative future Chinese AGI, the US military is already actively deploying advanced AI models in real-world combat scenarios. In Iran, for instance, US forces have reportedly utilized LLMs such as Claude and Grok to aid in planning bombing campaigns. Disturbingly, there are allegations that AI-assisted targeting may have played a role in the Shajareh Tayyebeh girls’ school bombing, an incident that tragically resulted in the deaths of over 150 elementary students and staff. This stark contrast between the hypothetical, future threat from China and the immediate, tangible ethical dilemmas posed by current US military AI use underscores a profound double standard. The focus on a speculative ‘boogeyman’ distracts from the urgent need for ethical frameworks and accountability in the deployment of AI in warfare today.
Stokes’ report, therefore, presents a perilous vision of future conflict, one driven by a blend of technological anxiety and geopolitical rivalry. It advocates for pre-emptive military action based on a highly theoretical future technology, while simultaneously sidestepping the very real, immediate challenges and ethical quandaries posed by AI in contemporary warfare. As the US and China navigate the complex terrain of AI development, proposals for kinetic strikes against critical infrastructure risk not only a devastating military escalation but also a dangerous precedent for international relations, pushing the world closer to a technological arms race with potentially irreversible consequences. Responsible AI governance, rather than speculative warfare, remains the most prudent path forward.
More on Chinese AI: US Companies Are Realizing That Chinese AI Models Are Way Cheaper, Ditch American Ones
