U of T researchers demonstrate AI worm could target any online device (www.utoronto.ca)

Researchers at the University of Toronto have unveiled a new class of cyberthreat: an AI worm capable of independently targeting and infiltrating any online device. Utilizing publicly available AI models, this malware can adapt its strategies as it spreads throughout networks, posing a significant danger to critical infrastructures like hospitals and financial systems. The research team, led by Nicolas Papernot, demonstrated that even those with limited resources could deploy sophisticated attacks using basic AI models, effectively hijacking devices to gain control and launch further strikes at minimal cost. This discovery challenges the current cybersecurity landscape, as traditional defenses may not adequately prepare for such adaptive threats. The AI worm operates by exploiting vulnerabilities in interconnected systems, learning and evolving with each infection, which renders static defenses ineffective. Papernot emphasizes the urgency of addressing this emerging threat through public awareness and improved security measures, urging organizations and individuals to strengthen their cybersecurity protocols. The researchers are concurrently developing countermeasures, highlighting the need for a collaborative approach across academia, industry, and policy-making to ensure the safety of the digital ecosystem.