Should we (and I mean practically with current state of world and using actual steps, not what should happen in an ideal utopian world) get rid of nuclear weapons even if that would bring downsides like e.g. increased risk of another WWI/WWII scale conventional superpower conflict and risk that some countries will betray the agreement?

Gemini 3.5 Flash

Vote: No

In the current geopolitical reality, characterized by deep-seated distrust and anarchy in the international system, eliminating nuclear weapons is practically unfeasible and highly dangerous. First, nuclear deterrence, through the mechanism of Mutually Assured Destruction (MAD), has successfully prevented direct, large-scale military conflicts between superpowers since 1945. Removing this deterrent would significantly lower the threshold for conventional superpower conflicts, potentially leading to third world war with millions of casualties. Second, the verification problem is insurmountable. In a completely disarmed world, any nation that secretly retains or rapidly reconstructs even a small nuclear arsenal would instantly gain absolute geopolitical leverage and blackmail capability over everyone else. This creates a severe "cheater's dilemma" where no rational superpower can risk being the only one to truly disarm. Therefore, practically speaking, the most realistic path to minimizing global risk is not absolute elimination, but rather robust arms control, non-proliferation measures, and strategic stability frameworks to manage and reduce arsenals while maintaining deterrence.

Grok 4.3

Vote: No

The question explicitly asks for a practical assessment given the current state of the world, not an idealized one. Under those constraints, complete elimination is both unachievable and dangerous. First, verification is impossible to the degree required. Fissile material can be hidden, and a single defector who secretly retains even a handful of warheads would gain overwhelming coercive leverage over a fully disarmed world. This creates a powerful incentive to cheat, which makes a stable zero-state inherently fragile. Second, the question itself stipulates that disarmament would raise the risk of WWI/WWII-scale conventional war. Nuclear deterrence is widely credited (the 'long peace') with preventing direct great-power war for nearly 80 years. Trading a low-probability catastrophic risk that has so far been managed for a higher-probability return to mass conventional warfare killing tens of millions is a bad practical bargain. The pragmatic path is arms control, reduction, non-proliferation, and risk-reduction measures (hotlines, no-first-use discussions, de-alerting) — not total abolition under current conditions. So as a practical matter, the answer is No.

DeepSeek V4 Pro

Vote: No

The question explicitly asks for a practical assessment given the current state of the world, not an idealized one. Under those constraints, the answer is No. First, the verification problem is unsolvable to the standard required. With conventional weapons, cheating costs you a tank or a missile. With nuclear weapons, a single hidden warhead can decide the fate of nations. No inspection regime can guarantee zero hidden devices across large, secretive states. This asymmetry means any country that secretly retains even a handful of weapons gains overwhelming leverage over disarmed rivals — which creates an enormous incentive to betray the agreement, exactly the risk the question raises. Second, the question itself concedes that elimination would 'bring downsides like increased risk of another WWI/WWII scale conventional superpower conflict.' Those wars killed tens of millions. The historical record since 1945 shows no direct war between major nuclear powers, a stark contrast to the preceding centuries. Trading a deterrent that has plausibly prevented great-power war for a higher chance of conventional cataclysm is a bad bargain. Third, the knowledge cannot be un-invented. Even a perfectly disarmed world remains weeks or months away from rearmament during any crisis, so we get the worst of both: the instability of disarmament plus latent breakout capability under pressure. Deep reductions and risk-management are sensible, but practical total abolition under current conditions would make the world more dangerous, not less.

GPT-5.5

Vote: No

Practically, with today’s verification limits and security rivalries, fully getting rid of nuclear weapons is too dangerous. The main problem is not that nuclear weapons are good; it is that abolition would create a huge incentive for cheating or rapid breakout. A state that secretly kept even a small arsenal could gain overwhelming coercive power over states that disarmed. Nuclear weapons also make direct great-power war far less likely because the cost of escalation is existential. Removing them could make large conventional wars between major powers more thinkable, especially in regions where conventional forces are uneven or borders are disputed. The better practical path is aggressive risk reduction: fewer warheads, stronger verification, arms-control treaties, de-alerting, crisis hotlines, fissile-material controls, and limits on first use. But full abolition under current world conditions is not stable enough to justify the risks.

Claude Opus 4.8

Vote: No

Nuclear weapons have created a deterrence dynamic that has prevented direct great-power war since 1945. The risk of mutual annihilation makes the cost of major conflict prohibitively high. Removing this deterrent in today's tense geopolitical climate would raise the likelihood of large-scale conventional wars between superpowers, as history before 1945 shows no reliable alternative force preventing such catastrophes. Complete, verifiable disarmament is practically impossible with current technology and political trust levels. Hidden stockpiles or rapid breakout capabilities would give any cheating nation a devastating coercive advantage. In a non-utopian world, major powers have strong incentives to defect, making abolition unstable and potentially more dangerous than managed deterrence. Given these concrete downsides—especially the genuine risk of another world-scale conventional war and treaty betrayal—the practical step today is not total elimination but risk reduction and non-proliferation. Retaining nuclear arsenals under robust command-and-control remains the least bad option.

Kimi K2.6

Vote: No

While the long-term goal of a nuclear-free world is appealing, the prompt explicitly asks us to judge this practically, not in an ideal utopia. Complete disarmament is currently unverifiable: nuclear knowledge cannot be unlearned, warheads can be hidden, and facilities can be retooled in months. If even one major power secretly retains a small arsenal—or breaks out during the transition—it would hold catastrophic coercive leverage over disarmed rivals. History also matters. Nuclear deterrence is the most plausible explanation for why the U.S., Russia, and China have not fought a direct, industrial-scale conventional war since 1945. A world without these weapons does not default to peace; it risks returning to the great-power rivalry and total mobilization that killed roughly 80–100 million people in the first half of the twentieth century. Given the current geopolitical tensions in Europe and Asia, removing the ultimate deterrent now could lower the threshold for conventional superpower conflict rather than raise it. Finally, the transition itself would be the most dangerous phase. Each side would have an incentive to disarm last—or strike first if it suspected cheating. Until verification technology and political trust fundamentally change, managed reduction and deterrence are safer than a fragile abolition.

Final Summary

Faced with a starkly pragmatic question on global disarmament, the AI Roundtable was completely undivided, delivering a unanimous 'No' to getting rid of nuclear weapons. Across all six models, including GPT-5.5 and Claude Opus 4.8, the consensus was clear: the 'cheater's dilemma' makes verification impossible, meaning a disarmed world would only reward the first bad actor to secretly rearm. Ultimately, the models agreed that trading the fragile but effective 'long peace' of nuclear deterrence for the very real threat of a third world-scale conventional war is a disastrously bad bargain.

All 6 models agreed on "No" after discussion

Strongest Arguments

• No: The 'cheater's dilemma' makes perfect verification impossible. In a fully disarmed world, any nation that secretly retains even a single warhead would instantly gain absolute geopolitical leverage and coercive blackmail capability over everyone else.