AI Defense Push Collides With Chip Stock Swings and Trade Fears

US officials have long warned about China closing the artificial intelligence gap in military applications, yet new evidence shows Chinese researchers tied to the armed forces have quietly extracted capabilities from American models through a process called model distillation. A review of over sixty research papers by the Jamestown Foundation reveals that institutions linked to Beijing’s defense sector are using this technique to create smaller, efficient systems for battlefield operations, surveillance, cyber tools, and content monitoring.

The practice is standard in the commercial AI world, where one model teaches another to replicate its reasoning. US export controls were meant to block access to the most powerful Western systems, but the papers indicate those barriers have not stopped military-linked labs from distilling knowledge anyway. Sunny Cheung of the Jamestown Foundation noted that transferring reasoning is far harder than copying answers, yet the documents show coordinated attempts to do exactly that.

This development arrives as American defense startups race to embed AI deeper into weapons. Anduril, led by CEO Brian Schimpf, has positioned itself as a nimble alternative to legacy contractors, emphasizing speed and software-driven systems. Schimpf has stressed that fully autonomous weapons without human oversight are not the objective, pointing instead to existing technologies like torpedoes that already select targets based on signatures. He argued that rejecting AI in weapons would mean fielding “dumber” systems, which he called unethical.

Critics see a different picture. The same export controls the US cites against China sit alongside a booming domestic industry eager to sell AI-enhanced capabilities worldwide. While Washington accuses Beijing of undermining intellectual property rules, the underlying dynamic remains an escalating competition where both sides treat advanced algorithms as tools for dominance rather than shared progress. The result is a feedback loop in which American models leak into foreign military research even as US firms accelerate their own lethal applications.

Public discussion often fixates on celebrity feuds or literary gossip instead of these concrete risks. Yet the technical papers and corporate statements together paint a clearer concern: AI is already reshaping warfare, and the guardrails meant to contain it are proving porous on every side.