ChatGPT Suits Up for the Pentagon
The day's clearest signal of where military AI is heading came from OpenAI, which plans to launch ChatGPT inside the U.S. Department of Defense in July, running on the Pentagon's GenAI.mil platform. It is a striking step: a consumer-famous chatbot moving into the Defense Department's own generative-AI environment. The arrival of commercial models in government workflows also raises the stakes on basic security hygiene — underscored today by a separate case in which a defense contractor settled False Claims Act allegations that it failed to implement required cybersecurity protections for sensitive information. The message to vendors is blunt: promise the safeguards, then actually build them.
Governments Move to Lock In — and Rein In — Military AI
Governance was the day's loudest theme. France formally designated military artificial intelligence an "essential security interest," a legal move meant to speed the technology into the hands of its armed forces. The parallel question — not how fast, but how restrained — drove the rest of the conversation. In Geneva, the UN Institute for Disarmament Research (UNIDIR) opened a forum on governing military AI, with a push toward actual enforcement rather than voluntary principles.
Industry is echoing the caution. Armin Papperger, CEO of German arms maker Rheinmetall, called for international guardrails and a firm line: humans, not algorithms, should decide when weapons fire. That tension runs through fresh analysis arguing AI could strengthen strategic deterrence — making an attack look too costly to attempt — while warning the underlying risks don't simply disappear when you add automation.
A more radical governance idea came from Capitol Hill. Senator Bernie Sanders filed a bill that would hand the American public a direct financial stake in the U.S. AI industry, with provisions that would put domestic AI firms partly in public hands. Whatever its odds, it reframes AI as national infrastructure rather than purely private enterprise.
The Chip Chokepoints Keep Tightening
None of these ambitions matter without silicon, and today the supply chain was full of friction. ASML — the only company on earth making the EUV lithography machines behind the most advanced chips — pushed back hard against a U.S. government report suggesting one of its tools had been shipped to China, calling the reports "damaging" and denying them outright.
Russia illustrated the flip side of that dependence: Moscow is mounting a fresh drive for AI "sovereignty" but keeps colliding with a basic wall — it cannot manufacture the advanced chips it needs. China, meanwhile, is tightening scrutiny of its indium exports as AI-chip demand surges. Indium rarely makes headlines, but it's a quiet input in chipmaking, and Beijing's grip on it is another lever in the materials standoff.
Europe managed to trip over its own feet. An "overly complex" EU legal provision could, by some readings, force Nvidia out of the bloc's AI buildout — a reminder that regulation, not just rivals, can stall a continent's ambitions. And in Washington, Treasury officials are weighing new curbs on American investment in China's biotech sector, extending the logic of tech containment beyond chips into the life sciences.
The Bottom Line
The through-line today: AI is being absorbed into the machinery of defense faster than the rules around it can set. Commercial models are walking into the Pentagon, France is fast-tracking adoption, and lawmakers are floating everything from public ownership to enforceable global norms. But every one of these moves rests on a hardware base that remains contested — over who builds the machines, who controls the metals, and who is allowed to buy in. Deterrence, sovereignty, and oversight all sound decisive on paper. Whether they hold depends on the chips, and on whether a human stays in the loop when it counts.