Would China engage in a MAIM dynamic?

2026 06 30

This essay was a 100 minute exercise thinking through whether China may engage in a MAIM dynamic as detailed in Superintelligence Strategy and its followup

Summary

China is unlikely to pursue systematic hard sabotage such as kinetic strikes or overt cyber disruption against rival advanced AI development in the near-to-medium term. Its current strategy favors economic statecraft, intelligence operations, and structural consolidation of its own AI infrastructure over the disruption strategy detailed by the MAIM report. If Chinese leadership develops awareness of MAIM-style dynamics, it would prefer to occupy the retaliator rather than initiator position given its existing development posture. The primary uncertainty is whether Chinese leadership comes to believe ASI represents an irreversible, winner-take-all shift in power that overrides the considerable strategic advantages of restraint, which remains open for now.

Dissecting the MAIM Framework

The Superintelligence Strategy framework holds that any state perceiving a rival approaching unilateral AI dominance will be motivated to preemptively sabotage that development—through cyber operations, kinetic strikes on data centers, and supply chain disruption. Coercive deterrence as detailed in this framework requires that the threatened pain appear contingent on the rival’s behavior. Moreover, it requires the threatened party believe compliance offers a genuine path out, and that the coercing party maintain credible willingness to act if resisted against. For MAIM to function as a stabilizing dynamic rather than a source of escalation, both sides should share a rough understanding of where thresholds lie and what crossing them would invite.

China fits several MAIM prerequisites for capability. Its leadership has demonstrated sustained strategic awareness of AI’s national significance. Its intelligence services have demonstrated willingness for aggressive technology acquisition across military and civilian domains. Its pre-positioning of cyber capabilities against critical infrastructure, documented in the Volt Typhoon operations, indicates China has already staged disruptive options without activating them. However, having the capability is separate from adopting a MAIM strategy.

China’s Belief in ASI

China’s current posture in its public and institutional discourse does not reflect strong belief in ASI as a winner-takes-all breakthrough. Chinese AI investment and policy is oriented toward applied capability, industrial deployment, and competition at the frontier; resource flows do not indicate heavy investment into an intelligence-recursion scenario, which is a critical threshold in MAIM. This is not to say industrial policy in China may not structurally induce a frontier capabilities breakthrough, but rather, with regards to the sabotage question: if China does not believe ASI confers decisive, monopoly-level strategic advantages, then it also will not perceive American frontier AI development as an existential threat requiring sabotage. The MAIM prerequisite that one perceives the rival is approaching a winner-takes-all capability may not be fulfilled under current perceptions.

China’s Economic Leverage

If China wished to deter American AI development, its most credible tools are not overt cyber strikes on data centers but economic pressure applied across the full breadth of the bilateral relationship, not just in AI-related commodities. China holds substantial leverage in the US-China bilateral trade relationship, through control over rare earths and mineral supply chains, and manufacturing dependencies that remain despite partial decoupling. Any trade pressure China can apply to the US functions as deterrence against American AI development, because it raises the overall cost of a confrontational posture. China’s economic leverage, thus, suggests China can deter development without disrupting. A Chinese decision to escalate AI competition to hard sabotage, on the other hand, would risk triggering economic decoupling and allied coalition-building that would most damage Chinese AI development prospects, forfeiting the geopolitical leverage China currently holds. The recent Beijing diplomatic engagement in May 2026 is also consistent with this reading, given both states demonstrated awareness of mutual economic dependency and preference for managed competition over destabilization. Deterrence through entanglement is China’s preferred mode, not through the threat of disruption.

The Strategic Retaliator Position

In a more speculative frame, consider the circumstances from China’s perspective once awareness of MAIM-style dynamics has emerged, that is, a general understanding that the state which visibly initiates sabotage bears significant political and strategic costs while the rival state gains legitimacy to retaliate. Both the US and China would structurally prefer to occupy the retaliator position, the party that responds to aggression rather than initiates it.

A state that strikes first against a rival’s AI infrastructure must justify that action domestically, to allies, and to non-allied states whose support it values. A state that retaliates against a demonstrated first strike faces a far lower justificatory burden. For China specifically, whose international positioning rests on multipolarity, non-interference, and sovereign equality, the retaliator position is especially valuable. If the US strikes Chinese AI infrastructure first, China gains an extraordinarily useful geopolitical narrative: the supposed defender of the rules-based international order is the aggressor, validating China’s longstanding critique of American hegemonic behavior. This framing would be effective across the Global South and potentially within American domestic politics itself. China’s posture of strategic patience would be disjoint from the US, where domestic politics and alliance obligations create pressure to act preemptively when a threat is perceived as sufficiently critical, as in preemptive war doctrine and forward defense.

This suggests China has an incentive to set up a provocation trap, making its own acceleration appear strategically threatening to compel a US first move, and its existing AI development already provides the conditions for retaliator legitimacy without needing to construct them artificially.

China’s deep integration of AI across critical civilian sectors—such as manufacturing, financial systems, urban infrastructure, and healthcare—further raises the cost of any American sabotage attempt by ensuring that degrading Chinese AI capabilities means degrading civilian and economic infrastructure. This is structurally analogous to Chiang Kai-shek’s move to place his most valuable troops in Quemoy, the US decision to place Western troops in West Berlin, or Xenophon positioning his own army with their backs against a ravine, each group unable to evacuate under fire. One places one’s most valuable assets in a cornered position where the adversary is required to choose between conflict and concession, making the adversary bear the moral weight of the choice. China, thus, may be compelled to take a strategically passive stance in the precursor to a MAIM dynamic.

The Limits of China’s Restraint

The above analysis holds as long as Chinese leadership does not come to believe ASI represents a decisive, irreversible winner-take-all shift in power. China currently has substantial economic assets embedded in a global order it values. However, should China, the less-resourced party, come to believe in the irreversible impact of ASI, it may conclude its fundamental strategic position, its autonomy, is being permanently foreclosed. China, then, may forego its normative commitment to multipolarity under existential pressure, accepting the costs initiating disruptive action to stall a permanent loss of power. The operative question here, then, is whether China will arrive at that belief, though the current evidence suggests this is not yet the case.

What China Will Do

Ultimately, China’s AI competition strategy will emphasize aggressive intelligence collection against frontier AI projects, covert model weight acquisition (hard to defend against), talent acquisition through legitimate and covert channels, acceleration of domestic semiconductor manufacturing and alternative compute pathways to reduce vulnerability to Western supply chain leverage, and applying economic statecraft across the US-China trade relationship. However, China will not pursue hard sabotage in the near term, due to its affinity for a more advantageous retaliator position. China’s best move, if it understands the coercive game being played, is to ensure it never has to be the one who shoots first. If a MAIM dynamic should be avoided, the Western community should similarly moderate Chinese expectations such that they never feel sabotage is worth the cost of their current geopolitical standing and domestic development.

Limitations

This is a brief memo inferring largely from my existing policy context on China. Extending upon the work, one should study the history of Chinese cyber operations and the scope of China’s actual trade leverage more carefully through data analysis, so as to stress test the economic deterrence claim.

Additionally, in the absence of a mutual understanding as detailed in the “Dissecting the MAIM Framework” section, the more subtle implication is that any sort of R&D sabotage compels the rival to invest more into AI security and safety regardless of whether they can attribute the failure to an adversary, thus reducing investment in frontier development and slowing progress. I was unable to deeply engage with implications for Chinese action in light of this, and would like to do so if given more time.

Currently, Chinese AI firms are engaging in harsh domestic competition, creating pressure toward frontier research, perhaps dangerously so. However, based on the historical pattern of Chinese industrial development, as seen in telecommunications, e-commerce, and fintech, where intense competition was followed by consolidation into a small number of state-favored national champions, I assume the current Chinese frontier AI landscape will simplify over time. Adversarially, this creates fewer targets for foreign sabotage, making intelligence gathering and sabotage easier. However, I haven’t investigated the full implications of this development either, and would prefer this be prioritized in future exploration.