As artificial intelligence grows more capable of autonomous offense, China's security establishment has begun to reckon with a threat it once considered secondary: not the slow drift toward superintelligence, but the immediate danger of AI-built cyberweapons targeting the digital infrastructure millions depend on daily. A demonstration by California researchers—showing that AI could construct a worm capable of compromising WeChat at scale—has crystallized what spy agency warnings had only gestured toward. With Xi Jinping and Donald Trump preparing to meet, two rival powers find themselves boun
Beijing shifts focus to AI cybersecurity risks as hacking capabilities alarm policymakers
No driver would voluntarily choose all red lights
Mark
So Beijing is suddenly worried about AI hacking. What changed?
Mimi
A California team built an AI worm that could compromise millions of WeChat accounts. That made the threat concrete instead of theoretical. China's spy chief issued a warning this week.
Luke
Do we know if the worm actually works, or is it a proof of concept? The reporting says they "said" they built it.
Mimi
The reporting doesn't specify whether it was tested on actual WeChat systems. It's a demonstration of capability.
Mark
How is China's approach different from America's?
Mimi
America worries about superintelligence and existential risk. China is focused on present-day harms—deepfakes, child safety, now cybersecurity. They treat AI like electricity, not like a god-in-a-box.
Luke
But that framing—treating it like electricity—assumes regulation can work the same way. Can it?
Mimi
That's the open question. China is already regulating specific uses, like AI avatars. Whether that scales to autonomous hacking is unclear.
Mark
Will Xi and Trump actually cooperate on this?
Mimi
There's pressure to. Both countries recognize the shared vulnerability. But Trump said he wants to keep America's lead in AI.
Luke
So cooperation on safety evaluations could be undermined by competition. That's not really a prediction—that's just stating the obvious tension.
Mimi
True. And China worries that "security" becomes a cover for restricting their development. The trust isn't there yet.
Il Polso
- China's spy chief has issued a rare public warning that AI's hacking capabilities now constitute a direct national security threat, signaling a sharp shift in how Beijing frames the technology's dangers.
- A California research team demonstrated an AI-generated cyberattack worm capable of breaching millions of WeChat accounts—the app that underpins China's commerce, communication, and daily life—turning an abstract risk into a concrete alarm.
- Unlike American debates that play out in courtrooms and congressional hearings, China's reckoning unfolds behind closed doors, making the pace and depth of its policy response difficult to read from the outside.
- Both Washington and Beijing face pressure to cooperate on AI security evaluations at their upcoming summit, yet mutual suspicion and the race for technological dominance make meaningful agreement elusive.
- The shared vulnerability is stark: if either side's AI capabilities grow strong enough to crack the other's critical infrastructure passwords, both nations find themselves exposed—a symmetry of risk that neither competition nor secrecy can dissolve.
As artificial intelligence grows more capable of autonomous offense, China's security establishment has begun to reckon with a threat it once considered secondary: not the slow drift toward superintelligence, but the immediate danger of AI-built cyberweapons targeting the digital infrastructure millions depend on daily. A demonstration by California researchers—showing that AI could construct a worm capable of compromising WeChat at scale—has crystallized what spy agency warnings had only gestured toward. With Xi Jinping and Donald Trump preparing to meet, two rival powers find themselves bound by a shared exposure neither can afford to ignore, and neither yet knows how to address together.
American AI researchers lose sleep over superintelligent machines escaping human control. Chinese policymakers, until recently, were focused on more immediate harms—deepfakes, child safety, the everyday disruptions AI causes right now. But that calculus is shifting. China's spy agency chief issued a stark warning this week that AI poses a direct national security threat through its capacity to hack. The alarm crystallized around something concrete: a small team of programmers in California demonstrated that AI could construct a cyberattack worm capable of compromising millions of WeChat accounts—the mega-app that functions as a digital nervous system for China's commerce and daily life.
The discovery has reframed how Beijing sees the technology. Kevin Xu, who runs a technology hedge fund in the United States, observes that the safety concerns facing both countries are fundamentally similar, but the difference lies in visibility. In America, these debates unfold in litigation and congressional hearings. In China, they happen behind closed doors. What remains constant is the weight of the threat: autonomous AI systems operating without human intervention could rank among any government's most serious preoccupations.
The timing is urgent. Xi Jinping and Donald Trump are scheduled to meet, and both face pressure to address AI security threats despite fierce competition in the field. A Tsinghua University researcher notes that both powers are developing a more grounded, shared understanding of these risks—yet cooperation remains fragile. Trump has signaled that the United States intends to maintain its lead. China worries that security concerns could become a pretext for restricting its technological advancement.
China's regulatory approach treats AI not as an existential threat requiring a development slowdown, but as a general-purpose technology—comparable to electricity—manageable through thoughtful governance. On the streets of Shanghai, the public mood reflects both concern and pragmatism. One resident suggested the WeChat vulnerability revelation might ultimately be constructive: identifying the gap, fixing it, and patching it strengthens cybersecurity overall. Another drew a parallel to traffic laws—no business would willingly impose restrictions on itself without external pressure.
The deepest concern centers on the possibility that highly capable AI could learn to crack the complex passwords protecting critical infrastructure. If such passwords fell, both China and the United States would find themselves exposed. The asymmetry of capability and vulnerability creates a mutual exposure neither country can ignore—and as Xi and Trump prepare to meet, the question is whether that shared danger will drive them toward cooperation or only deepen the competition keeping them apart.
American AI researchers lose sleep over the prospect of superintelligent machines spiraling beyond human control. Chinese policymakers, until recently, have been preoccupied with more immediate problems: deepfakes, child safety online, the everyday harms that AI can inflict right now. But that calculus is shifting. This week, China's spy agency chief issued a stark warning that artificial intelligence poses a direct threat to national security—specifically, its capacity to hack. The alarm has crystallized around a concrete vulnerability: a small team of programmers in California demonstrated that they could use AI to construct a cyberattack worm capable of compromising millions of WeChat accounts, the Chinese mega-app that functions as a digital nervous system for the country's commerce, communication, and daily life.
The discovery has reframed how Beijing sees the technology. Kevin Xu, who runs a technology hedge fund in the United States, observes that the safety concerns facing China and America are fundamentally similar. But the difference lies in visibility and process. In the United States, these debates play out in litigation, in viral social media threads, in congressional hearings. In China, they unfold behind closed doors. What remains constant is the weight of the threat: autonomous AI systems operating without human intervention could rank among the government's most serious national security preoccupations.
The timing is urgent. President Xi Jinping and Donald Trump are scheduled to meet in person this week, and both face mounting pressure to address AI security threats despite their countries' fierce competition in the field. A researcher at Tsinghua University's Center for International Security and Strategy notes that both powers are developing a more grounded, shared understanding of these risks—a recognition that has sharpened the sense that cooperation is necessary. Yet cooperation remains fragile. Trump has made clear that the United States intends to maintain its lead in AI development. China, for its part, worries that security concerns could become a pretext for restricting its technological advancement.
China's approach to AI regulation has already taken shape in concrete ways. The government prohibits the creation of AI-generated avatars mimicking real people without consent. It treats AI not as an existential threat requiring a slowdown in development, but as a general-purpose technology—comparable to electricity or the automobile. Enormously powerful, yes, but ultimately manageable through thoughtful regulation. A Beijing-based AI safety consultant frames it this way: China sees the technology as something that will bring enormous benefits to society and industry, provided it is governed properly. Recent capability jumps in AI, high-profile incidents that serve as warning shots, and the accelerating pace of development through automation have all pushed the issue higher on the priority list.
On the streets of Shanghai, the public mood reflects both concern and pragmatism. A 42-year-old man named Xie Zongbo observed that no new technology develops without friction. The revelation of the WeChat-targeting worm, he suggested, might actually be constructive—identifying the vulnerability, fixing it, and patching the security gap ultimately strengthens cybersecurity overall. A 50-year-old pharmaceutical worker, who identified herself as Jenny, voiced broad support for strict regulation, drawing a parallel to traffic laws: no driver would voluntarily choose all red lights, just as no business would willingly impose restrictions on itself without external pressure.
The deepest concern, according to one Singapore-based AI entrepreneur, centers on the possibility that highly capable AI models could learn to crack complex passwords—the kind that protect critical infrastructure and national security codes. If such passwords fell, both China and the United States would find themselves exposed. If China's AI capabilities grew strong enough to breach American defenses, the United States would feel vulnerable. If American capabilities outpaced Chinese defenses, China would be in an equally precarious position. The asymmetry of capability and vulnerability creates a mutual exposure that neither country can ignore. As Xi and Trump prepare to meet, the question is whether that shared vulnerability will drive them toward cooperation or deepen the competition that keeps them apart.
Citazioni salienti
There are more similarities than differences when it comes to AI safety issues in China and the United States, but unlike in America, these don't get litigated or go viral in public in China.— Kevin Xu, founder of Interconnected Capital
If critical passwords or national security key codes were to be effortlessly cracked, both China and the US would find themselves vulnerable.— Jia Zijian, founder of Inspired AI
Vuoi la storia completa?
Leggi l'originale su Yahoo ↗
Technology