Scientists just built a powerful AI computer worm that learns as it spreads

Researchers have created an AI-powered computer worm designed to spread autonomously between devices, demonstrating a cybersecurity threat that could outpace existing defenses. The study, posted on arXiv and reported by the *New York Times*, used an undisclosed but freely accessible AI model—avoiding proprietary systems like those from Anthropic or OpenAI—to build the prototype. Unlike traditional malware, this worm can learn and adapt as it spreads, targeting hidden vulnerabilities in networked systems. The findings highlight a critical gap in global cybersecurity preparedness, as modern infrastructure—including water, energy, healthcare, and financial systems—relies heavily on interconnected computers. David Lie, a cybersecurity expert at the University of Toronto, emphasized the need for rapid development of countermeasures, calling the research a ‘wake-up call.’ The worm was tested in an isolated virtual environment, but its potential for real-world exploitation remains a major concern. The study’s authors stress that AI-driven malware poses unique risks, as it can evolve beyond pre-programmed instructions to exploit new weaknesses. While AI itself could also help mitigate such threats, the dual-use nature of the technology means both offensive and defensive applications must advance urgently. Experts warn that without proactive measures, the consequences of AI-powered cyberattacks could be catastrophic for global infrastructure.