University of Toronto researchers have demonstrated that AI-driven "worms" could autonomously target and infect online devices. By leveraging open-weight large language models (LLMs) to identify vulnerabilities and execute attacks, these agents can spread across networks by exploiting security flaws. The project highlights a concerning reality: AI can now automate the entire lifecycle of a cyberattack, from reconnaissance to self-replication. While some experts argue this is an inevitable progression of existing malware techniques, others warn that the shift toward autonomous, self-evolving agents represents a significant escalation in risk. Discussion on Hacker News reflected a mix of alarm and skepticism. Critics questioned the novelty of the research, noting that the underlying concepts are already well-understood by security practitioners. Conversely, proponents argue that public demonstrations are essential to forcing industry-wide improvements in network security and internal defenses. The conversation underscores a broader anxiety about the "AI-driven" future: as these models become more capable, the barrier to creating sophisticated, self-spreading cyber threats continues to lower. Despite the debate over the research's novelty, there is a consensus that current internet infrastructure is poorly prepared for the rise of autonomous, intelligent malware.