AI-built WeWorm hijacks WeChat without a tap

23rd September 2026 – (Palo Alto) A Palo Alto cybersecurity firm says it used artificial intelligence to build WeWorm in little more than a week — a zero-click worm that can seize WeChat accounts from an unanswered voice call and then spread through a user’s contacts.

Researchers at Calif found that an attacker need only place a WeChat voice call to the target. The victim does not have to answer, tap a link or touch the phone: within seconds the flaw can fire and the account is taken over. Once inside, an attacker can read chat histories, send messages and place calls, impersonating the real user across the same social graph that once defined trust.

Worse, the hijacked account then dials contacts automatically, propagating like a worm between Apple’s iOS and Android. Even hanging up does not stop the attack. Calif calls it the first known computer worm able to jump platforms without any victim click or touch — a break from older malware that needed a careless download or a tapped message.

Tencent has confirmed the vulnerability and says it fixed the issue after Calif’s notice, so users need not update the app themselves. Calif says the remediation was completed on the server side on 28th August, closing the path for the demonstration exploit without a forced client upgrade for ordinary handsets still running older WeChat builds.

Zhao Minghao, deputy director of Fudan University’s Center for American Studies, said that given WeChat’s ubiquity in China, cracking the app carries destructive force akin to “a new kind of nuclear weapon”. The platform spans daily messaging, government services and digital payments, he argued, so disabling or mass-compromising it would hit livelihoods and public administration at once.

Brookings Institution researcher Chen Kaixin said the episode underscores the risk of non-state actors using AI against China’s digital infrastructure — and may push Beijing to seek tighter cooperation with Washington on managing AI hazards before similar tools leave the lab.

Compound AI threats are widening fast. Model-maker Anthropic recently said it had thwarted scientists trying to use its systems for biological-weapons work, while Microsoft co-founder Bill Gates and other technology leaders warn that containing AI risk is a global priority. That agenda is expected to feature among key topics when the US and Chinese leaders meet later this month, as capitals weigh how quickly research demos can become real-world pressure on everyday apps.