AI Becomes National Security Weapon

Close-up of a circuit board with an AI symbol and digital effects

China is quietly weighing new limits on sending its most powerful AI systems overseas, raising fresh fears that global technology is being locked behind national security walls.

Story Snapshot

The United States has already blocked foreign access to some advanced AI models using national security powers, setting a clear precedent for export controls.
Those U.S. controls briefly forced Anthropic to shut off its Fable 5 and Mythos 5 models worldwide, including for foreign workers inside the company.
Evidence shows China takes AI risks seriously, but there is still no public Chinese order that copies Washington’s export bans.
Growing limits on AI trade deepen the sense, on left and right, that elites are turning critical technology into a weapon of power instead of a tool for ordinary people.

What the U.S. just proved it can do to AI exports

When the United States Department of Commerce ordered Anthropic to cut off foreign access to its Fable 5 and Mythos 5 models, it showed how far modern export controls can reach. The directive did not target a physical chip leaving a port. It targeted an online service and model weights that lived in the cloud. The order forced Anthropic to shut down access for “any foreign national,” even those working inside its own offices. For many businesses, that felt less like trade policy and more like a kill switch.

Trump administration used national security law after officials flagged a jailbreak that could bypass safety guardrails and help launch cyberattacks. That narrow technical flaw became the legal hook for a sweeping global ban. Commentators later noted that export controls once focused on hardware like high-end NVIDIA chips, but are now shifting to model weights and even real-time application programming interface access. In other words, Washington has learned how to choke off brains, not just the brawn, of advanced AI.

Why China now sits in the crosshairs of AI controls

Analysts describe a steady campaign by the United States to limit China’s access to the building blocks of artificial intelligence. Rules announced in 2022 stopped American companies from selling top-tier AI chips and key chip design tools to Chinese buyers. Later measures added dozens of firms to blacklists and extended controls to cloud-based computing and AI diffusion. One research group bluntly said the strategy aims to “strangle the Chinese AI industry” by cutting off the parts needed to run and train advanced systems. That kind of language feeds suspicion on both sides that export controls are now a geopolitical weapon.

At the same time, lawmakers in Washington are trying to wall off Chinese models inside the United States. A bipartisan bill called the No Adversarial AI Act would bar federal agencies from using AI built in China, Russia, Iran, or North Korea. The proposal responds to claims that China’s DeepSeek system helps its military and intelligence services. For many Americans, this looks like common sense security. For others, it proves elites are deciding which tools people can use, with little public debate and even less transparency.

What we know, and do not know, about China’s own stance on AI exports

Unlike the detailed U.S. orders, there is still no primary-source Chinese document in public that explicitly calls for blocking Chinese AI models from overseas markets or cutting off foreign users. The available research shows concern in Beijing about AI risks, not clear rules about exports. A major Chinese policy framework warns that open-source models could help criminals train “malicious models” and even help build weapons of mass destruction. Chinese experts now discuss extreme risks, such as AI systems becoming uncontrollable. That debate proves China is thinking hard about safety, but it does not yet confirm an export clampdown.

The US just unblocked Anthropic's most powerful models — Fable 5 and Mythos 5 — for global release. Here's what actually happened.

Three weeks ago, the Trump administration flagged both models as national security risks and slapped export controls on them. Today, Commerce…

— Teksart (@TeksCreate) July 8, 2026

Security researchers describe China’s approach to AI as strongly state-directed, treating AI as a strategic public good tied to national power. They contrast that with the United States, where government mostly lets private companies drive AI growth. If China does move toward strict export controls on its own models, it would fit this pattern of tight central control. Yet current reporting offers no signed order from China’s Ministry of Commerce, no leaked meeting notes, and no court cases that show a formal ban on sending Chinese AI systems abroad. For now, talk of Chinese export limits rests on inference, not hard proof.

How growing AI walls feed public anger at the “deep state”

Across the political spectrum, many Americans already feel the federal government serves the powerful more than the public. The new AI export battles reinforce that fear. Think about Anthropic’s case: a little-known directive, based on a technical finding that ordinary users never saw, instantly cut off access for millions and even threatened to delay a public stock offering. Similar chip controls have wiped out tens of billions of dollars in U.S. company value and cost jobs, all decided by small groups of officials and lawyers. That looks to many like rule by insiders, not by voters.

At the same time, warnings about China’s AI—surveillance tools, military applications, and systems like DeepSeek—are real and serious. Lawmakers and security experts argue that if the United States lets rivals tap American-made AI too freely, it could help build weapons or spy systems turned back against American citizens. That core tension drives today’s policy chaos: people want safety from hostile regimes and rogue code, but they also fear their own government using “national security” as a blanket excuse to control what tools they can access. If China now considers its own export limits, it will only add another layer to this growing wall around AI.