Researchers Build WeChat Worm Hijacking Phones via Zero-Click Calls

"A worm at this scale used to be the kind of thing that took a larger team months. AI can already do most of the work here." — Calif researchers

What the researchers built: WeWorm and its capabilities

A Palo Alto startup called Calif says it has built a tool, dubbed WeWorm, that can hijack WeChat accounts and seize control of phones simply by placing a call. According to a disclosure dated September 8, Calif's proof-of-concept gives an attacker full control of a targeted WeChat account: the attacker can read and send messages, make calls and act on the victim’s behalf. The team described WeWorm as “the first zero-click worm to spread through WeChat calls across iOS and Android.”

The technical root: a VoIP memory corruption bug in WeChat

Calif reported that the underlying vulnerability is a memory corruption issue in WeChat’s voice-over-IP (VoIP) stack. The flaw, the researchers wrote, leverages the privileges that WeChat grants to trusted contacts during communication. They said the bug allows remote code execution (RCE) when exploited during an incoming call, and that the victim need not answer or otherwise interact with the device for the exploit to succeed.

Timeline, testing, and vendor response

The researchers said they discovered the bug in July using a mix of large language models (LLMs), including both open-weight models and closed-source models from “US frontier labs”; they declined to identify which models. Calif reported the flaw to Tencent, and said their WeChat account was initially banned after that report. Tencent later confirmed that exploitation could allow remote command execution and released patched app versions: Android 8.0.77 and iOS 8.0.76.

Calif reported that they produced functional exploits for vulnerable WeChat apps in two days and spent another week integrating those exploits into the WeWorm tool. They tested the tool on several lab phones, specifically citing Google Pixel 10a models and an iPhone 17e model.

How an attack works — and its practical limits

Calif’s disclosure stresses two operational points. First, the exploit is “zero-click”: the victim does not need to pick up the phone, and if they do answer they reportedly hear nothing while the exploit still succeeds. Second, the attacker must be on the victim’s WeChat friend list for the call to trigger the flaw. The researchers called that constraint “not much of a barrier,” noting that an attacker could first compromise one friend’s account and then use that account to reach additional targets. They also warned that, when chained with other Android and iOS bugs, WeWorm could lead from account takeover to full device control.

Development context: AI-assisted discovery and rapid weaponization

Calif explicitly credited large language models with accelerating their research workflow. They said LLMs helped them find the vulnerability and that AI “can already do most of the work here,” while their team provided judgment about what to target and how to test exploits safely. According to Calif, the combination of AI-enabled discovery and a small, focused team allowed the construction of a worm-style tool in a matter of days—work that used to require larger teams and months of effort, in their view.

What this means for technologists, end users, and adversaries

• Technologists and security teams: The patched WeChat app versions named by Tencent are Android 8.0.77 and iOS 8.0.76; teams responsible for mobile fleets and app security will need to confirm those builds are deployed to remove the specific RCE vector Calif described. Calif also warned that exploitation can be extended by chaining other platform bugs, so mobile defenders should be alert to combined attack paths.
• End users and account owners: Because the exploit operates over incoming WeChat calls and does not require user interaction, users cannot rely on ignoring calls alone; however, the researchers noted that actively declining an incoming call stops that particular attempt. The friend-list requirement means account compromise of trusted contacts is a plausible prerequisite for wider spread.
• Adversaries and threat actors: Calif’s demonstration shows a pathway for rapid weaponization of a messaging-app RCE into an account-hijacking worm that spans iOS and Android. The tool as described grants attackers the ability to act on victims’ behalf within WeChat once the call-based exploit succeeds.

Calif’s disclosure offers a concrete example of a voice-call vector combined with automated tooling to produce a cross-platform worm that targets a widely used messaging app. Tencent’s issuance of patched versions provides a clear remediation path for the exact flaw Calif reported — Android 8.0.77 and iOS 8.0.76 — while the researchers’ account of how quickly exploits were created underlines how short the window between discovery and weaponization can be.

https://www.infosecurity-magazine.com/news/wechat-zeroclick-worm-hijack/