Google DeepMind researchers have issued a stark warning about a new class of cybersecurity threats targeting autonomous AI systems, as agents increasingly evolve beyond passive chatbots into active web-interacting entities. SSRN March 2026 paper highlights that the study introduces “AI Agent Traps,” a novel attack technique that exploits how AI agents perceive, interpret, and act on […] The post Hackers Can Hijack AI Agents via Malicious Web Content, Warn Google DeepMind appeared first on Cyber Security News.