AI and national security: attacks turned 'warning shots' and put Beijing on alert

The ability of advanced artificial intelligence models to conduct hacker attacks has become a matter of national security in China. The head of the country's espionage agency has publicly warned that technology can compromise national security, and researchers have demonstrated a worm capable of reaching millions of WeChat accounts. According to Japan Today, episodes act as "warning shots": they are not yet catastrophes, but show what lies ahead.

Why did China change the speech about AI and national security?

So far, the Chinese debate on AI revolved around content: deepfakes, misinformation and labelling of synthetic material. The novelty is the shift to national security and cyberdefense — a framework that usually comes with more state control and more investment in detection. The heads of espionage alert signaled that Beijing began treating border models as potential offensive weapons, not just as productivity tools. It is not an isolated movement: similar alerts came from Western agencies, as the Five Eyes alert on cyber attacks with AI shows.

What does WeChat's worm reveal about AI attacks in practice?

Researchers presented a worm — self-replicating code that spreads itself across systems — capable of compromising millions of WeChat accounts, China’s leading messaging app. The central point is not the number itself, but the scale: a committed account can turn thousands without human intervention. It is the same pattern described in research on AI models that invade computers and self-replicate alone. Classic worms depended on known failures; when a generative model helps to find loopholes, the propagation rate changes.

Why are AI attacks called warning shots for national security?

The expression summarizes the reading of the authorities: the incidents recorded so far did not overturn critical infrastructure or paralyze services, but prove that capacity already exists. Treating each case as a sample — and not as an exception — is what justifies classifying the risk as national security. In practice, this tends to mean mandatory verification of models, restrictions on autonomous invasion testing tools, and monitoring of anomalous traffic on mass platforms.

What changes for companies and developers in Brazil with AI and national security?

Whoever uses AI agents in production inherits a similar attack surface. API keys, broad permissions, and automatic execution of generated code are the natural way for a worm to spread within a company. For Brazilian teams, the practical lesson is to treat the model as an unreliable component: limit the scope of credentials, review call logs and test propagation scenarios before someone does. Regulators also tend to look at the subject from the perspective of security, not just privacy.

What is still unclear about AI and national security

The source does not detail which models were used in the tests, whether the worm was corrected, or whether the affected platform was pronounced. There is also no indication of concrete measures already in force in China because of the alert. Without this data, the episode functions more as a driving sign than as a damage balance.

Frequently Asked Questions about AI and national security

What did China say about AI and national security?

The head of the country’s spy agency warned that advanced AI models can compromise national security, according to Japan Today — a shift of focus, which was previously content and misinformation.

What's WeChat's worm?

It is a self-replicating code demonstrated by researchers, capable of compromising millions of WeChat accounts, China's leading messaging app.

Does this affect Brazilian companies?

Indirectly. Companies running AI agents with broad credentials are exposed to the same type of automatic propagation, which reinforces the need to limit permissions and monitor calls.