Geopolitics & War Brief — July 23, 2026

The Big Story

For 18 days in June, two of America's most advanced AI models simply vanished from the internet. On June 12, 2026, the U.S. Commerce Department ordered Anthropic to stop providing its Fable 5 and Mythos 5 models to any foreign person without a special license ^[1]. The company faced an impossible choice: obtain licenses for every foreign user, or shut the models down entirely. It chose the latter. Public access to Fable 5 was restored on June 30 ^[1].

The episode revealed a bigger problem than any single model. The United States does not have a trusted, standardized process for evaluating the security risks of frontier AI models or proportionate remedies when those risks are identified ^[1]. What happened to Anthropic was less a carefully calibrated policy decision and more an emergency shutdown—the digital equivalent of a circuit breaker with no dial. The government ordered it, the company complied, and millions of users lost access to a tool they relied on, all without the public ever fully understanding why.

This matters for every investor watching the AI sector, because it means the regulatory environment isn't just opaque—it's improvised. A government can order a shutdown with 18 days' notice, and no one knows what comes next. That kind of uncertainty creates risk that no amount of quarterly earnings growth can price in.

What Else Moved

Chinese Spies Are Already in the Room

Former U.S. officials told the House Intelligence Committee this week that foreign intelligence agencies—particularly China—are actively targeting American AI labs, and that U.S. agencies have not kept pace with the threat ^[6]. The message was clear: while policymakers debate whether to restrict AI exports, adversaries are stealing the technology outright. This raises a uncomfortable question: are export controls like the Anthropic case addressing the real vulnerability, or are they security theater while the actual threat walks through the front door?

Lawmakers War-Game a Taiwan Nightmare

Representatives participated in a tabletop exercise hosted by CSIS this week in which they weighed responses to simulated Chinese cyberattacks on U.S. infrastructure during a China-Taiwan conflict ^[4]. The scenario wasn't abstract. These attacks were AI-enabled, meaning adversary software was making decisions in real time without human oversight. For lawmakers accustomed to thinking about military hardware—ships, jets, missiles—the exercise forced a reckoning: the next major war could start with code, not carriers. And America's defense posture, built around visible military assets, may not be ready.

Australia's Drone Breaks Through (Quietly)

Boeing's MQ-28 collaborative combat aircraft (a drone designed to work alongside piloted jets) flew in a U.S.-led Pacific exercise despite not winning the formal U.S. competition for collaborative combat aircraft ^[3]. A Boeing executive called it an important "first step." The quiet victory suggests that even as geopolitical competition intensifies in the Indo-Pacific, military partnerships are advancing through smaller incremental steps. Australia's participation signals that allied nations are willing to integrate new autonomous systems into joint operations—a shift that would have been unthinkable a decade ago.

Canada Signals Alignment on Defense Tech

Canada's defense minister this week reinforced a call for increased cooperation among middle powers ^[5]. The comment came amid Canada's decision on future fighter jet procurement (the GCAP program), but the real subtext is broader: allied nations are repositioning to hedge against Chinese and Russian technological advantage. Canada's moves, like Australia's drone integration, suggest a coalition-building moment in Western defense strategy.

Connecting the Dots

Three urgent truths emerge from this week's stories. First: the U.S. government has regulatory authority over AI but no coherent policy. It can shut down models but cannot explain why—or prevent spies from stealing the same technology anyway ^[1][6]. Second: the next military threat is arriving faster than the institutions designed to counter it can adapt. Lawmakers are just now learning what AI-enabled cyberattacks look like in a Taiwan scenario; meanwhile, China's spies are already in American labs ^[4][6]. Third: allied nations are moving ahead without waiting for Washington to figure it out. Australia is flying new drones in U.S. exercises; Canada is signaling tighter defense alignment ^[3][5]. They cannot afford to wait for American policy to crystallize.

The pattern is one of asymmetry. Adversaries are stealing, integrating, and advancing. Allies are adapting. And America is improvising governance one crisis at a time.

What to Watch

Watch for the Commerce Department's next move on frontier AI models. Will the Anthropic case become a template, or was it a one-off ^[1]? Second, monitor how the China-Taiwan war game shapes congressional funding and oversight of defensive AI and cyber capabilities ^[4]. Third, track whether Australia's drone integration in U.S. exercises expands to other allied nations—a sign that the West is building interoperable autonomous systems ^[3]. Finally, listen for what the House Intelligence Committee does with the warning about Chinese espionage in AI labs; if it leads to new funding or security protocols, the threat has reached critical mass ^[6].