Washington has spent years debating the position of synthetic intelligence (AI) in nuclear command and management, and agonizing over the army implications of China-U.S. AI competitors. It has spent far much less time asking a distinct query: how will North Korean chief Kim Jong Un’s deterrence calculus change if he believes the USA is on the verge of acquiring a strong, enduring benefit in synthetic normal intelligence?
Synthetic normal intelligence (AGI) – outlined as highly effective intelligence matching human efficiency throughout cognitive duties – might additional destabilize nuclear deterrence relationships which can be already deteriorating. Weaker, non-AGI powers might turn into particularly problematic. North Korea’s nuclear posture is already on a hair set off. If the USA obtains AGI or AGI-like capabilities, Pyongyang might undertake an aggressive “use it or lose it” mentality as a result of it fears a primary strike that eradicates its nuclear weapons capabilities and decapitates its management.
Critically, the USA’ precise acquisition of AGI and the following growth of deterrence-nullifying capabilities issues lower than North Korea’s perceptions of U.S. capabilities and intent. As a result of Pyongyang has adopted an aggressive nuclear posture and espouses such a vital, paranoid view of U.S. intentions, the scenario might spiral even earlier than AGI capabilities are operational. Extra consideration ought to subsequently be paid to how AGI might upset uneven deterrence equilibria, particularly vis-à-vis Pyongyang.
AGI Takeoff Shouldn’t Be Dismissed
If AGI takeoff actually unlocks exponential progress, a cascade of technological breakthroughs might spur unprecedented financial progress and growth in army capabilities that some have termed agentic warfare. Underneath this situation, an AGI-capable energy might quickly erode, if not remove, the second strike capabilities of a non-AGI energy. This technological leap would amplify the unlucky psychological tendency to turn into extra risk-prone when dealing with losses, known as prospect principle.
Accordingly, underneath AGI takeoff, non-AGI nuclear weapons powers, particularly the weaker ones, will possible worry a near-future whereby they are going to be unable to discourage the AGI energy. They may subsequently undertake a extra aggressive posture. After all, this isn’t the one rational response, however for causes we’ll describe later, we expect North Korea is more likely to reply with additional aggression than it’s to hunt refuge by survivability, concealment, and arms management negotiations. Furthermore, if recursive self-improvement delivers synthetic superintelligence (ASI), outlined as capabilities exceeding any human, the stakes will turn into a lot increased and nuclear decision-making timelines will turn into much more compressed.
AGI penalties, timelines, and diffusion stay a topic of debate, and key AI breakthroughs could also be a enterprise cycle or extra away, however the steadiness of proof suggests large functionality will increase are attainable. If scaling legal guidelines proceed to carry, capabilities will maintain rising: introduced U.S. AI computing capability is projected to extend by 239 % this yr alone, based on Federal Reserve employees calculations based mostly on Epoch AI information. By 2030, deliberate U.S. AI compute could be roughly 35 occasions mid-2025 ranges.
Whereas tech CEOs and traders are incentivized to hype AI advances, fashions are undeniably bettering, and at an spectacular charge; the most recent fashions are even proving adept at working within the bodily world. For higher and for worse, AI software program and {hardware} capabilities are advancing quickly. Even synthetic intelligence specialists themselves have expressed shock on the tempo of innovation and warned of the hazards of unrestricted growth.
Strategic deterrence nullification would possible lag AGI by a few years, however AI timelines are advancing far sooner than most anticipated. North Korea’s pressure posture is already altering and new command-and-control safeguards – together with human-in-the-loop commitments – could take years to agree upon and implement.
AGI attainment would usher in appreciable geopolitical disruptions, together with altering the nuclear balance-of-power. China and Russia are of particular significance in any AGI situation. China is a peer competitor of the USA, together with in AI, and it possesses the world’s third largest nuclear arsenal; Russia possesses the world’s largest nuclear weapon arsenal (by quantity) and continues to face off in opposition to the USA in Chilly Battle-style proxy battle. Both competitor might be incentivized to behave aggressively if the U.S. is first to strategy AGI attainment.
Nonetheless, the gravest AGI-related risks could come from less-capable nuclear powers, particularly North Korea. That’s as a result of Pyongyang has adopted a uniquely aggressive and decentralized type of nuclear command and management, has an aggressive nuclear doctrine, suffers from extreme technical and even geographic disadvantages, and it maintains a paranoid disposition in the direction of the USA. Confronted with the prospect of shedding second strike functionality, it might subsequently select to undertake an excellent riskier posture or operationalize half-baked tech into its nuclear command, management, and communications earlier than the USA crosses the AGI threshold.
Nuclear stability requires mutual vulnerability. China and Russia usually tend to proceed to carry the U.S. hostage through risk of retaliation even when the U.S. obtains AGI first; for North Korea, that’s a lot much less possible.
AGI Takeoff, or Perceived Takeoff, Upends Standard and Strategic Army Calculations
If AGI takeoff happens, its technological, financial, and army penalties will probably be extraordinarily laborious to forecast. Bodily constraints and {hardware} bottlenecks, together with in vital minerals, might gradual the transformation of software program breakthroughs into weapons. Nonetheless, highly effective AI might quickly ship unprecedented financial and technological capability, which might then translate into army capabilities, together with early warning, detection, focusing on, and supply. These capabilities might credibly threaten different states’ second-strike forces.
As the choice window shrinks, mutual vulnerability – the muse of a secure deterrence relationship – will erode. Given the hazards of dealing with an AGI-empowered rival, and their very own weak monitoring and verification capabilities, non-AGI powers like Pyongyang will watch anxiously for indicators that an adversary’s AGI attainment might unlock the potential for a decapitation strike that wipes out their management and nuclear capabilities in a single fell swoop.
On this situation, dealing with quickly deteriorating strategic choices, a state might be motivated to automate nuclear retaliation. They may hope that the useless man’s hand would function a ultimate guarantor of deterrence. However as human operators are pulled additional out of the nuclear decision-making course of, the moral and sensible issues will start to multiply. Historical past has confirmed that people are the final finest protection in opposition to unnecessary escalation. That was the case in 1983 when a Soviet early warning system erroneously signaled an incoming U.S. intercontinental ballistic missile. Utilizing his information of the defective system and U.S. technique, Lieutenant Colonel Stanislav Petrov deduced that the warning was a false optimistic. He was proper: the system’s satellites misinterpret daylight reflecting off clouds.
As well as, North Korea – and different non-AGI actors – would possibly search to ascertain deterrence through different uneven strategies, equivalent to pre-positioning chemical and organic weapons or creating new, AI-enabled variants. North Korea is believed to have a substantial stock of each chemical and organic weapons, and it would lean on these capabilities if it fears its nuclear deterrent is compromised.
To restrict near-term organic dangers from Pyongyang, each Washington and Beijing ought to carefully look at the AI-biology nexus, an asymmetrical risk that North Korea would possibly discover compelling to develop if it believed a rival was on the cusp of AGI acquisition. Limiting the chance of North Korea – or different state and non-state actors – could require constraining the publicly-accessible organic capabilities of future releases of huge language fashions. Unsafeguarded fashions, often called open-weight fashions, pose essentially the most extreme dangers. As unsafeguarded fashions turn into extra succesful, just about anybody can entry highly effective, unconstrained AI by eradicating safeguards with comparatively little compute.
North Korea is already possible working unsafeguarded fashions to take advantage of zero-day vulnerabilities, because the $6 billion Binance hack suggests. Given the rising dangers of AI-designed viruses, together with from Pyongyang, Beijing and Washington ought to have candid conversations concerning the desirability of diffusing highly effective AI capabilities in biology.
North Korea Is Uniquely Weak and Aggressive
In contrast to China or Russia, which have massive nuclear arsenals, mature intelligence, surveillance, and reconnaissance capabilities, and demonstrated technological potential to innovate, North Korea’s vulnerability and its aggressive posture could render it essentially the most harmful nuclear actor in an AGI dash situation. Moreover, in contrast to China or Russia, international locations that may depend on second-strike deterrents as a result of their deep interiors and nuclear-armed submarines, North Korea lacks a credibly survivable nuclear submarine deterrent.
Lastly, North Korea’s personal eagerness to undertake AI-enabled weapons will shade its notion of adversary intent and the risk posed by AGI. The nuclear deterrence equilibrium on the Korean Peninsula is already fragile; inviting yet another destabilizing ingredient could be the straw that breaks the camel’s again.
North Korea’s army is embracing AI with a velocity that means it’s throwing warning to the wind. In 2024, the Basic Workers Division of the Korean Individuals’s Military created an Digital Warfare and AI Operation Command to develop and undertake “AI-based digital warfare assist programs.” In August 2025, the Korean Individuals’s Military issued orders to its Strategic Power to “combine AI-based unified administration programs for storing, working, and commanding nuclear weapons, in addition to launching nuclear counterattacks,” based on Every day NK. On the Ninth Get together Congress in February of this yr, North Korea introduced extra plans to combine AI into its civil and army programs. Since then, much more developments have been introduced.
Chief Kim Jong Un himself jumped behind the wheel of a cellular rocket launcher, claiming that the nuclear-capable 600-mm a number of rocket launcher (KN-25) options synthetic intelligence in its steering programs. One month later, Kim oversaw a check of AI-enabled assault and reconnaissance drones, saying, “unmanned tools and synthetic intelligence needs to be top-prioritized… in modernizing the armed forces.” In Could, North Korean media claimed that the army examined a tactical cruise missile that makes use of synthetic intelligence for goal acquisition and terminal steering. Regardless of this urge for food for integrating AI throughout army applied sciences, limitations in {hardware}, information, and infrastructure will possible constrain the rollout. These limitations might additionally produce defects that imperil safeguards meant to cease an unintentional escalation spiral.
Additional aggravating the scenario, North Korea’s nuclear doctrine is uniquely aggressive and opaque. The nation’s rubber stamp parliament handed laws in 2022 that legitimizes preemptive nuclear assaults underneath sure circumstances. The North will launch a nuclear assault “robotically and instantly” if Kim’s command is threatened, if “a deadly army assault” on vital targets is imminent, or if wanted to reclaim “the initiative in warfare.” The legislation additionally outlined a delegative command authority: “Quite than await direct orders from Kim Jong Un, discipline commanders could now comply with predetermined launch plans to launch nuclear weapons.” These doctrinal reforms have been strengthened by a 2026 constitutional revision that codified the decentralized command authority. Mixed with a surge within the manufacturing and ahead deployment of tactical nuclear weapons, these adjustments add coherence and credibility to Kim’s first-use technique, which goals to attain a most deterrence impact by aggression and ambiguity. Nonetheless, the blurry pink strains and threats of preemption are dangerous and escalatory.
An AGI situation might result in additional brinksmanship. As Keir A. Lieber and Daryl G. Press famous in a 2017 evaluation, “The New Period of Counterforce,” states dealing with rich, technologically superior adversaries will discover it “more and more tough to safe their arsenals” as advances in sensors, information processing, and AI erode the circumstances which have traditionally protected second-strike forces.
Pyongyang can’t rule out that an AGI-powered United States will develop next-generation applied sciences, equivalent to airborne directed power weapons, extremely efficient anti-ballistic missile protection, and orbital and sub-orbital weapons. Pyongyang’s weak early warning and intelligence capabilities will amplify the management’s worry that its second strike has been rendered moot, particularly given its distinctive vulnerabilities. Nor can Washington simply reassure North Korea: even when the USA integrates AI cautiously into its personal nuclear command, management, and communications and is clear about human-in-the-loop safeguards, a management this distrustful might conclude that the AGI second has arrived and act accordingly.
It’s attainable that, confronted with such daunting odds, North Korea might select a path of survivability, restraint, and even reconciliation. That might be a rational response to this newfound unfavorable place. Nonetheless, given Pyongyang’s nuclear posture, its historic document, and the management’s constant dedication to sturdy nuclear capabilities, it’s more likely that Pyongyang will double down on its “escalate to de-escalate” doctrine by adopting even riskier and extra aggressive measures.
Fearing a super-capable U.S. and preoccupied with the specter of a decapitation strike, Pyongyang might be extra prone to settle for the downstream dangers that come from including AI-enabled processes. The primary effort could be into their very own nuclear command, management, and communications structure, together with early warning programs, focusing on, command, management, communications, and navigation. Given North Korea’s already aggressive and dangerous posture, this embrace of AI can solely improve the probability that an unintended conflict might spiral right into a full-on nuclear change. As well as, any assault on the North Korean infrastructure supporting the Korean Individuals’s Military AI digital warfare programs might be interpreted as a nuclear-level aggression, resulting in a nuclear-level reprisal.
The efficacy of AI programs is restricted by the standard of coaching information, sensory enter, and processing energy. In North Korea’s case, the standard of all three is extraordinarily restricted. This might produce false positives, which turn into exponentially extra harmful as people are faraway from the decision-making course of. Pairing super-compressed resolution timelines with automated assault protocols makes it tough for a cool-headed human to steer the scenario away from disaster. That’s doubly the case when the 2 states in query lack diplomatic relations, disaster communications, or confidence constructing measures.
Opposite to widespread depictions – together with Annie Jacobsen’s guide “Nuclear Battle: A Situation” and Kathryn Bigelow’s 2025 movie “A Home of Dynamite” – North Korea’s management understands effectively that launching a pre-emptive strike could be suicidal. However the alarming fact is that Pyongyang might make rather more delicate and obscure posture adjustments that will dramatically improve the prospect of unintended escalation. Going through an AGI-enabled United States, North Korea could be more likely to rapidly operationalize half-baked AI nuclear command, management, and communications, hedge with chemical and organic weapons, ahead deploy extra property, have interaction in grey zone assaults, and have interaction in aggression that blurs the road between nuclear testing and use, equivalent to an atmospheric check close to Korea or Japan that produces no casualties. They might additionally broaden the record of pink line objects that set off discipline commanders to independently launch a nuclear strike.
The Want for (Considerate) Pace
The US faces a fancy optimization downside: staying forward of Chinese language AI growth whereas managing the nuclear deterrence instability that AGI might unleash, and the potential existential dangers of AI itself. Policymakers ought to start inspecting and counteracting these dangers now, as an alternative of scrambling in a takeoff situation. Constraining Chinese language entry to superior semiconductors is critical however not ample. U.S. policymakers should additionally think about how their very own AGI growth will have an effect on deterrence calculus in different capitals and discover voluntary restraints on capabilities that will destabilize fragile nuclear equilibria.
The following frontier of arms management ought to embody a “human within the loop” clause such because the one featured within the 2024 settlement between the USA and China. The U.S. ought to lead a multilateral effort to safe purchase in from all nuclear powers to decide to the identical provisions and extra – to wall off as a lot higher-function nuclear command, management, and communications from automaticity as attainable. Whereas it might appear counter-intuitive to contemplate that Pyongyang would settle for limitations on its nuclear program, participation in such a multilateral settlement with the world’s nuclear powers would possibly truly bestow upon it the form of worldwide legitimacy that Kim craves. And this kind of legitimacy is categorically completely different from prior choices as a result of it will situate North Korea amongst nuclear superpowers on an equitable foundation, a tantalizing prospect for Kim that can’t be simply dismissed.
Whereas some could criticize such an effort on the grounds of making a harmful precedent by tacitly accepting Pyongyang’s nuclear standing, the time has come to cease tilting at “denuclearization” windmills in favor of reaching a cloth risk discount, as one in all us has argued beforehand. To make certain, convincing the distrustful management in Pyongyang to voluntarily settle for limitations on its nuclear program is not going to be a simple promote. However the prices of inaction are so excessive and the alternate options so unpleasant that, within the land of awful coverage choices, that is an endeavor worthy of focus, power, and political capital.
Beijing might play a job in containing nuclear-related and uneven organic dangers. Given the dangers of highly effective AI diffusing organic analysis capabilities, and enabling an uneven counter to AGI, each the U.S. and China have an curiosity in constraining AI’s publicly-accessible organic capabilities. The identical applies to limiting North Korea’s nuclear capabilities. As throughout the Six Get together Talks, Beijing needs to be recruited to assist Pyongyang’s engagement and implementation.
Over time, such a framework might start to eat away on the mistrust that fuels North Korea’s paranoia and spurs its aggressive nuclear posture. If Kim Jong Un continues to reject good religion efforts, then the U.S. and its allies South Korea and Japan ought to make investments considerably and collaboratively on missile protection structure and different capabilities wanted to deal with the risk and preserve deterrence. That consequence ought to incentivize China to behave responsibly.