The unveiling in January of China’s most sophisticated AI model marks a turning point. As President Trump cautioned in his recent speech on technological security, this development is a “” for American leadership. The stakes are not just about economic competition, but potentially the most geopolitically sensitive technology since the atomic bomb.
Following Oppenheimer’s creation of the atomic bomb, the U.S. held a technological advantage for roughly four years before the Soviet Union caught up. This balance of terror, coupled with the unprecedented destructive power of these weapons, led to the concept of mutual assured destruction (MAD). This deterrence strategy, while imperfect, prevented major conflict for decades. The threat of nuclear retaliation discourages a first strike, leading to a tense but stable equilibrium.
Today’s AI competition could be even more intricate than the nuclear era because AI is a versatile technology impacting various sectors, from medicine to finance to defense. Advanced AI could even automate AI research, giving the leading nation a growing advantage in both defense and offense. A nation nearing superintelligent AI, which surpasses human intelligence in almost every aspect, would pose a national security threat to rivals, potentially leading them to consider sabotage rather than concede power. If we are moving toward a world with superintelligence, we must acknowledge the potential for geopolitical instability. A newly released details the geopolitical implications of powerful AI and suggests a comprehensive “Superintelligence Strategy.”
Consider how the U.S. might respond if rival nations sought an insurmountable AI advantage. If Beijing were to surpass American AI labs and approach recursively improving superintelligence first, U.S. national security would be severely and fundamentally threatened, regardless of whether Beijing could control its creation. The U.S. might rationally resort to threatening cyberattacks on AI datacenters to prevent China from achieving its goal. Similarly, Xi Jinping—or Vladimir Putin, who is unlikely to develop the technology first—might react the same way if the U.S. neared recursively improving superintelligence. They would not passively accept an impending U.S. monopoly on power.
Just as the destabilizing pursuit of nuclear dominance eventually yielded to the stability of MAD in the nuclear age, we might soon face a similar deterrence dynamic with AI. If any nation attempting to achieve AI supremacy faces the threat of preemptive sabotage, nations might be discouraged from pursuing unilateral power. This outcome is what we term Mutual Assured AI Malfunction (MAIM). As nations recognize this possibility, we anticipate it becoming the standard approach, and we must prepare for this new strategic landscape now.
MAIM is a framework designed to maintain strategic advantage, prevent escalation, and curb the ambitions of rivals and malicious actors. To ensure its effectiveness, the U.S. must clearly communicate that any destabilizing AI project, especially those aimed at achieving superintelligence, will be met with retaliation. In this context, offense—or at least the credible threat of it—is likely the best defense. This involves expanding our cyberattack capabilities and increasing surveillance of adversary AI programs.
While constructing this deterrence framework, America must simultaneously progress on two additional fronts: AI nonproliferation and domestic competitiveness.
For nonproliferation, we should implement stricter AI chip export controls and monitoring to prevent compute power from falling into the wrong hands. We should manage AI chips similarly to uranium, maintaining precise records of product movement, incorporating restrictions on the capabilities of high-end AI chips, and empowering federal agencies to monitor and shut down illegal distribution channels.
Finally, to sustain a competitive advantage, the U.S. should prioritize building resilient supply chains for military technology and computing power. Specifically, our reliance on Taiwan for AI chips represents a significant vulnerability and a critical bottleneck. While the West currently holds a strong position in AI chips, Chinese competition could disrupt this. Therefore, the U.S. should enhance its domestic design and manufacturing capabilities. Superintelligent AI presents a challenge as complex as any policymakers have encountered. It is what theorists Horst Rittel and Melvin Webber termed a “wicked problem,” constantly evolving with no definitive resolution. MAIM, supported by strong nonproliferation efforts and renewed investment in American industry, offers a strategy based on lessons from past arms races. While there is no purely technical solution to fully control these forces, the correct balance of deterrence, nonproliferation, and competitiveness measures can help the United States navigate the emerging geopolitical reality of superintelligence.
“`