Rivalry over data and networks, not robot autonomy, magnifies the danger of escalation: vulnerabilities in sensors, communications and training data make attribution murky, weaken deterrent signals and speed decision cycles, increasing the risk of miscalculation between the United States and China.

Main Finding

The paper argues that U.S.–China competition over military AI primarily raises escalation risks through contested networks and datasets (cyber exposure), rather than through the autonomy of individual platforms. Action–reaction pressures and adoption incentives in autonomous weapons, ISR, and decision‑support tools reshape crisis stability by increasing operational uncertainty, undermining signal credibility, and accelerating or delegating decisions.

Key Points

• Central thesis: escalation risk stems mainly from dependence on contested data and networks (e.g., vulnerabilities to cyberattack, data poisoning, spoofing), not merely from platform autonomy.
• Three illustrative capability areas examined:
  • Autonomous weapons — adoption pressures, integration with networks, and implications for target attribution and accountability.
  • Intelligence, surveillance, and reconnaissance (ISR) — reliance on shared sensing and fusion increases opportunities for deception and denial.
  • Decision‑support tools — AI that accelerates or delegates decision cycles can shorten crisis timelines and amplify erroneous responses.
• Three mechanisms through which AI competition reshapes strategic interaction:
• Operational uncertainty and attribution ambiguity — AI-enabled systems make cause, source, and intent harder to determine, raising risks of misperception.
• Signal clarity and data integrity — contested data flows and manipulated inputs degrade the credibility of signals used for deterrence and reassurance.
• Decision acceleration and delegation — faster, automated decision loops compress escalation time horizons and increase the role of pre‑delegated responses.
• Focus on cyber exposure places information infrastructure, data governance, and secure networks at the core of strategic stability concerns.
• The paper draws on official doctrines, strategic plans, and policy assessments to structure the analysis (qualitative, document‑based approach).

Data & Methods

• Methodology: structured interpretive analysis of authoritative sources — official doctrines, strategic plans, and policy assessments from the United States, China, and relevant international bodies.
• Analytical approach: conceptual mapping of strategic mechanisms (uncertainty, signal integrity, decision timing) onto three capability domains (autonomous weapons, ISR, decision‑support).
• Evidence type: qualitative documentary evidence and doctrinal texts used to illustrate incentives, stated plans, and potential operational consequences.
• Limitations: no new empirical datasets or formal modeling presented; the argument is theory‑driven and interpretive, identifying plausible mechanisms rather than quantifying probabilities.

Implications for AI Economics

• Strategic value of data and secure networks: data, sensor access, and trusted communications become strategic economic assets; firms and states will compete over data control, provenance, and resilient architectures.
• Investment incentives and externalities: competition creates strong incentives for rapid adoption and integration of AI military capabilities, but also negative externalities (security races, underinvestment in verification and resilience).
• Market for cybersecurity and trusted AI: rising demand for robust cyber defenses, data‑integrity tools, provenance and authentication services, and secure hardware — opportunities for private-sector growth and government procurement.
• First‑mover vs. fragility tradeoffs: early adopters may gain operational advantages but also expose themselves to exploitation of immature systems and networks; economic models should incorporate vulnerability costs and insurance/contingency pricing.
• Dual‑use spillovers and industrial policy: military-driven AI demand will shape civilian AI ecosystems (talent, compute, data markets); policymakers may need targeted support for defensive R&D, secure data infrastructures, and controls on sensitive transfers.
• Governance and verification economics: international coordination, norms, and verification mechanisms are public goods that reduce strategic instability but face collective‑action problems; economic analysis can inform optimal mixes of export controls, investment screening, subsidies for defensive capabilities, and liability/insurance regimes.
• Modeling suggestions for researchers: dynamic adoption/arms‑race models with network externalities, endogenous data/control accumulation, cyberattack probabilities, and strategic signaling under information degradation would help quantify risks and policy tradeoffs.