American models, Chinese defense systems
The day's dominant story comes from Reuters: Chinese military researchers have used outputs from leading American AI models — including those built by OpenAI and Anthropic — to help train domestic AI systems intended to advance China's defense capabilities. Reuters reached that conclusion through its own review of the material.
The claim traveled fast and unevenly. A separate write-up from Roya News, surfaced through Google News, circulated the same basic assertion — that China is extracting outputs from American AI systems to train its own defense technology — in a thinner, second-hand form. Worth keeping the two straight: one is a documented review, the other is the echo of it. The underlying allegation is the same, and it is the one that matters.
The distillation front opens in parallel
Running alongside that story is a narrower, sharper accusation. The US government has accused Moonshot AI — the Chinese startup behind the Kimi K3 model — of illegally "distilling" Anthropic's advanced Fable model, according to reporting aggregated on MSN and surfaced through Bing News.
Read together, the two items describe the same anxiety at two different scales: the fear that America's lead in frontier models is not a moat at all if the outputs of those models can be harvested, refined and reused abroad. A model's weights can be locked down. Its answers, by design, cannot.
Meanwhile, Washington is trying to blacklist Anthropic
The irony of the day: while Chinese researchers were reportedly training on Anthropic outputs, the Pentagon was in federal court trying to brand Anthropic itself a supply chain risk — and losing ground.
At a hearing Thursday, a federal judge pushed back on the Trump administration's designation, saying the government still hasn't shown its work. Reporting from Gizmodo, Tech Times and Newsbytes describes a skeptical bench and an administration that has so far failed to put evidence behind the label. Nothing is settled, but the government's position is not landing.
Trade walls go up, and Beijing promises payback
The administration also moved to block new Chinese-made AI robots and power inverters from the American market, citing national security, according to a report published on Yahoo Finance. Beijing has vowed retaliation. Inverters are the unglamorous half of that pairing and arguably the more consequential one — they sit inside energy infrastructure, which is precisely why they've drawn scrutiny.
The backdrop, per The Guardian, is that China's rapid technological progress is scrambling assumptions on both sides of the American tech world — in Silicon Valley boardrooms and in Washington policy offices alike. Today's actions read like a system reacting to that realization in real time.
Contractors move deeper into classified work
Quieter but telling: Nvidia's partners are pushing further into US defense projects, including work that carries security classifications. HPCwire reports that a data storage company backed by Nvidia has gotten a foot in the door on classified AI. The commercial AI supply chain and the national security apparatus keep converging — which is exactly what makes fights like the Anthropic designation so consequential.
And in India: fifty drones, one mind
Finally, India's Defence Research and Development Organisation is pursuing a new generation of swarm drones, targeting a system in which as many as 50 drones fly and operate together as a single coordinated unit. Coordination at that scale is the hard part — and the whole point.
The through-line: every story today is about control. Who controls model outputs, who controls the supply chain, who controls the swarm. Nobody got a clean answer.