Generative A.I. models have been used to create a self-replicating computer worm designed to target WeChat, according to a new report from A.I. researchers. The malware is capable of breaking into user accounts and spreading on its own. Experts warn that the attack could potentially compromise hundreds of millions of devices within hours if deployed widely.
The worm was developed by a team of researchers who aimed to expose vulnerabilities in current A.I. systems. Unlike traditional malware, this worm leverages the capabilities of large language models to automate its hacking process. It can manipulate chat functions and payment systems within the WeChat platform, making it particularly dangerous.
The attack works by exploiting weaknesses in the A.I. models that power the application’s user interactions. Once a device is infected, the worm can scan contact lists and send malicious messages to other users. Those messages then activate on new devices, allowing the worm to replicate and spread without human intervention.
Security analysts who reviewed the research described the findings as a significant alarm for the tech industry. They noted that most existing security defenses are not equipped to handle threats that use generative A.I. to adapt their attack methods. The worm’s ability to quickly evolve its tactics sets it apart from conventional viruses or phishing scams.
WeChat, which has over a billion users globally, is a major communications hub, especially in China. The platform integrates messaging, social media, and digital payments into a single app. This broad functionality gives malicious software a larger attack surface to exploit, raising the stakes for potential data theft or financial fraud.
The researchers conducted their test in a controlled environment to contain the threat. Despite the sandboxed setting, the worm successfully spread across simulated user accounts faster than expected. This speed underscores the ease with which A.I.-driven malware could escalate from a single infected phone to a worldwide outbreak.
The tech community’s response has been mixed, with some praising the research for highlighting a real danger, while others question the ethics of building a functioning malicious tool. The creators have shared their methodology with select security firms to help patch vulnerabilities, but they have not released the full code publicly.
Cybersecurity experts are now calling for new regulatory frameworks to address the unique risks posed by A.I.-generated malware. They argue that traditional antivirus software is insufficient and that platform developers must redesign their systems to monitor for anomalous behavior at the A.I. layer. The report serves as a key reminder that the tools used to build smarter applications can also be weaponized.
For now, no active attacks using this worm have been detected in the wild. However, researchers advise users to enable two-factor authentication and be cautious about clicking unknown links. The full impact of the discovery will only become clear as more security teams analyze the potential countermeasures.
