Researchers Build WeChat Zero-Click Worm Hijacking Phones via Calls

The emergence of WeWorm highlights a paradigm shift in cybersecurity, where the barrier to entry for developing complex, cross-platform exploits is being dramatically lowered by the integration of large language models (LLMs). According to the disclosure report published by Calif on September 8, the development process for this worm—which would have previously required a team of specialized security engineers months of labor—was completed in a matter of weeks, with the core exploitation phase taking just two days.

The Mechanics of the WeWorm Vulnerability

At the heart of the security breach is a memory corruption flaw residing within the voice-over-IP (VoIP) stack of WeChat. The vulnerability allows an attacker to gain unauthorized access to a victim’s device without requiring any user interaction. In traditional phishing or social engineering attacks, a victim must click a malicious link, download a suspicious file, or grant permissions to an application. In contrast, WeWorm operates as a "zero-click" vector; the malicious payload is delivered via the signaling protocol of a standard WeChat call.

The victim does not need to answer the incoming call for the exploit to execute successfully. Even if the user ignores the notification, the underlying vulnerability is triggered by the arrival of the call packet itself. If the user does answer, they hear nothing, as the exploitation process occurs silently in the background while the application processes the call data. Declining the call effectively terminates the current attempt, but researchers noted that this is a temporary deterrent, as an attacker can simply initiate a new call attempt, for example, during the late-night hours when a user is less likely to be monitoring their device.

While the exploit requires the attacker to be present on the victim’s "friends list," the researchers emphasize that this is a negligible hurdle in a practical attack scenario. The worm is designed to propagate autonomously; once a single device is compromised, it can access the victim’s contact list and initiate calls to other users, effectively turning every compromised account into a node in an expanding web of infection.

AI-Accelerated Vulnerability Discovery

The methodology employed by the Calif research team underscores the increasing role of artificial intelligence in software vulnerability research. The team utilized a combination of proprietary, closed-source LLMs from major U.S. frontier labs and open-weight models to automate the identification of the memory corruption flaw.

The researchers have not disclosed the specific models utilized, nor have they provided granular technical details of the exploit itself to prevent malicious actors from replicating their work. However, they confirmed that the AI models were instrumental in analyzing the complex codebase of the WeChat VoIP stack to pinpoint memory management errors that would have been exceptionally difficult for a human auditor to identify manually.

This reliance on AI for offensive security research represents a double-edged sword. While it allows security firms to find and patch vulnerabilities faster than ever before, it also democratizes the ability to discover "zero-day" flaws. As LLMs become more proficient at reading and debugging complex codebases, the window between the disclosure of a vulnerability and the development of an active exploit is expected to shrink significantly.

Chronology of the Disclosure and Remediation

The discovery and subsequent remediation of the WeWorm vulnerability followed a structured timeline that highlights the importance of responsible disclosure.

• July 2024: The Calif research team identifies the memory corruption vulnerability in the WeChat VoIP stack using AI-assisted analysis.
• Late July 2024: The researchers initiate the standard disclosure process, reporting the findings to Tencent, the parent company of WeChat.
• August 2024: During the reporting phase, the researchers note that their own WeChat accounts were temporarily banned, a standard automated response from Tencent’s security systems when unusual activity or unauthorized testing is detected.
• Early September 2024: Tencent verifies the vulnerability, confirming that it could indeed lead to remote command execution. The company releases patches for both Android (version 8.0.77) and iOS (version 8.0.76).
• September 8, 2024: Calif releases their full disclosure report, detailing the existence of WeWorm and the potential implications of the exploit, while withholding the specific technical mechanics to ensure user safety.

The rapid development of the exploit—two days for the core vulnerability and one additional week for the integration into the WeWorm tool—demonstrates the high level of efficiency afforded by the team’s automated workflow.

Broader Impact and Implications

WeChat is far more than a simple messaging application; it is a "super-app" that integrates social media, mobile payments, ride-hailing, healthcare booking, and government services. With over a billion monthly active users, a vulnerability that grants an attacker "full control" over a WeChat account has profound consequences.

An attacker who successfully executes the WeWorm exploit gains the ability to read and send private messages, make unauthorized calls, and potentially manipulate financial transactions within the app. Furthermore, the researchers warned that the exploit can be chained with other, separate bugs within the Android and iOS operating systems to escalate privileges, potentially leading to full control of the host device. This means that a successful WeChat breach could act as an entry point for deeper surveillance, data theft, or the installation of persistent malware on the victim’s smartphone.

The ability of the worm to spread autonomously through contact lists creates a "viral" risk, where the infection can propagate through large social networks in a matter of hours. This is reminiscent of classic worm threats like "Morris" or "SQL Slammer," but updated for the mobile-first era and targeted at the most sensitive applications in a user’s digital life.

The Future of Mobile Security

The Calif researchers conclude that the era of manual, labor-intensive exploit development is nearing its end. Their project serves as a "proof of concept" for a future where AI handles the heavy lifting of security research, leaving human experts to focus on judgment, strategic targeting, and the ethical implications of their work.

For users, the primary defense remains the consistent application of software updates. The speed at which Tencent was able to push out a patch for the WeChat vulnerability is a positive indicator, but the incident underscores the vulnerability of even the most popular and "trusted" applications to sophisticated, automated exploits.

Cybersecurity analysts suggest that this event will likely prompt developers of other major messaging platforms—such as WhatsApp, Telegram, and Signal—to conduct deep, AI-driven audits of their own VoIP and signaling stacks. As the threshold for creating high-impact exploits drops, the security industry must evolve to match the speed and precision of AI-driven adversaries. The WeWorm incident is unlikely to be an isolated event; it is a harbinger of a new landscape where the integrity of global communications is perpetually tested by the very tools designed to secure them.

As users continue to rely on centralized super-apps for an increasing share of their digital activity, the concentration of risk within these platforms becomes a critical policy and technical challenge. Future mitigation strategies will likely need to move beyond simple bug patching and toward more robust, compartmentalized architectures that limit the blast radius of any single vulnerability, ensuring that even if a communication app is compromised, the integrity of the underlying operating system and the user’s broader data remains intact.