WeWorm Explained: The AI-Assisted Zero-Click WeChat Worm That Hijacked Accounts

WeWorm was an AI-assisted, proof-of-concept computer worm built by Palo Alto security firm Calif to demonstrate a serious flaw in WeChat’s voice-calling code. Calif said a specially crafted incoming call could compromise a recipient’s WeChat account in seconds, even if the person never answered or interacted with the phone. The company reported the issue to Tencent before public disclosure, and Tencent deployed mitigations. 1014

How the zero-click call attack worked

The underlying issue was a memory-corruption vulnerability in WeChat’s Voice-over-IP (VoIP) stack—the component that handles calls. Calif said the exploit could execute while a call was ringing, eliminating the usual need for a victim to click a link, open an attachment, or answer a call. 103

That made it a zero-click attack: the vulnerable software processed the call before any meaningful user action was required. Calif has withheld the low-level vulnerability and exploit details, an important safeguard against replication. 10

Why it could spread like a worm

A worm differs from a one-off account takeover because it can use each newly compromised account to reach further targets. In Calif’s demonstration, the compromised WeChat account could place calls to people in its saved contacts or friend list, attempting the same attack again. 110

This contact-based propagation was both powerful and constrained:

• It could exploit trust relationships: a call originating from a familiar account may appear less suspicious.
• It could move across WeChat’s contact graph without an attacker manually targeting every victim.
• It depended on an available contact/friend relationship, rather than reaching any arbitrary WeChat user from scratch. 1

Calif demonstrated a chain spanning Android and iOS devices, showing that the propagation concept was cross-platform. Claims that it could reach hundreds of millions of accounts were projections of potential spread through WeChat’s large network—not evidence that such an event occurred. 1013

What access an attacker would have gained

Calif described the result as full control of the WeChat account. That could allow an attacker to read and send messages, place calls, impersonate the account holder, and use the account to target additional contacts. 18

That is not the same as a complete iPhone or Android operating-system takeover. The demonstrated impact was account-level control; Calif indicated that separate vulnerabilities would be needed to turn that access into full device compromise. 14

What AI contributed—and what remains unknown

According to Calif, its team used AI to find the flaw and write an initial remote-code-execution exploit in about two days, then built the worm demonstration in roughly another week. 108

The key implication is not that AI independently launched an attack. Rather, the project illustrates how AI can reduce the time and specialist effort needed for vulnerability research, debugging, exploit iteration, and adaptation. Those same capabilities can also help defenders analyze software and accelerate remediation.

Calif’s public WeWorm account says the team worked with AI, but it does not reliably identify a particular model or coding product. Claims that WeWorm specifically used Claude, GPT, Codex, or another named tool are therefore not supported by the available public disclosure. 10

Disclosure, patches, and real-world risk

Calif said it privately reported the vulnerability to Tencent in July. Tencent released WeChat Android 8.0.77 and iOS 8.0.76 on August 21, according to Calif, and Calif later said a server-side mitigation blocked the exploit for all users. Tencent confirmed that the vulnerability could enable remote command execution. 1014

The distinction between a demonstration and an active campaign matters: public reporting describes WeWorm as a proof of concept, and there is no public evidence that Calif deployed it against real users or that it caused an in-the-wild mass infection. 9

The security lesson from WeWorm

WeWorm is notable because it combined three high-impact properties: zero-click exploitation, self-propagation through trusted contacts, and cross-platform reach. Even with the reported flaw mitigated, the project is a reminder that communications features—especially call handling and other automatically processed content—can become high-value attack surfaces.

For users, keeping WeChat and mobile operating systems updated remains the practical defense. For software teams, the broader lesson is to prioritize memory-safety testing, secure call-processing paths, rapid vulnerability disclosure channels, and defense-in-depth for account recovery and session controls. 1014