Researchers at the University of Toronto have discovered an AI-powered worm that can spread across computer systems without direct human control. The experiment has been conducted in a controlled setting, and the entire report is now out in a paper titled ‘AI Agents Enable Adaptive Computer Worms.’
The study has drawn attention from cybersecurity experts since it shows how artificial intelligence could change the way malware operates in the future. Here, the researchers have highlighted that a new generation of computer worms could use artificial intelligence to spread across networks autonomously. They don’t need human intervention; instead, they adapt their strategy when they encounter a challenge or roadblock.
According to researchers, the worm can move from one system to another by identifying vulnerabilities on its own. Once it gained access, it could continue searching for other vulnerable systems and attempt to spread further. Unlike traditional forms of malware, it did not need constant instructions from a human operator.
The research was conducted on systems running Linux, Windows, and the Internet of Things. The researchers have noted that within a week, the malware was spread to three-fourths of the devices without any human commands.
The study does not mean such worms are already spreading widely online. However, it highlights a risk that security teams may face as AI tools become more common and more powerful.
Also Read: Best Cybersecurity Audit Firms in 2026: Top 10 List
The findings have also prompted cybersecurity companies to accelerate the development of AI-based defense tools. Many firms are developing systems that can monitor network activity, detect unusual behavior, and respond faster than traditional security software. This is important because most current security systems were built to detect known threats. An AI-powered threat that can change its behavior may be harder to identify using older methods.
The new study suggests that the next phase of cybersecurity could involve AI on both sides, with intelligent attacks facing equally intelligent defenses.