New research from the University of Toronto has demonstrated an AI-powered worm capable of adapting its attack strategy as it moves through a network, raising concerns about the future effectiveness of traditional cybersecurity controls.
Researchers at the University of Toronto have published new research demonstrating an AI-powered worm capable of adapting its attack strategy as it moves through a network, highlighting a potential new challenge for cybersecurity teams.
Unlike traditional worms, which typically exploit a fixed vulnerability, the prototype can reason through new attack paths and tailor attacks to different systems. Researchers said this could make such threats significantly harder to stop using a single patch or signature-based defence.
The findings suggest that existing defensive models may face increasing pressure as AI enables malware to adapt and mutate while operating inside a target environment.
According to Ken Ammon, CEO of CodeHunter and a former US Air Force captain at the NSA, the development challenges many of the assumptions underpinning traditional malware detection.
“The significance of AI-powered worms is not that they spread automatically. Worms have existed for decades. The difference is that AI allows malware to adapt, mutate, and develop new attack paths as it encounters different environments. That undermines many of the assumptions behind traditional malware defenses, which depend on recognizing known patterns, signatures, indicators of compromise, or behaviors after execution has already begun,” said Ammon.
He said organisations will need to place greater emphasis on software trust and security earlier in the development lifecycle.
“The critical security question is no longer whether code looks malicious based on what we’ve seen before. It is whether that code should be trusted to execute at all,” said Ammon.
