In an age where AI like OpenAI's ChatGPT becomes increasingly capable, we are handing over mundane tasks to artificial agents — from scheduling calendars to shopping. But with the enhancement of such tools, the scope for cyber vulnerabilities proportionally increases. A recent demonstration by researchers at Cornell Tech pinpoints a novel threat: generative AI worms. These worms can autonomously propagate through systems, capable of stealing sensitive data or spreading malware. The gravity of this innovation is its unparalleled modus operandi in cyberattacks, undetected until now, according to researcher Ben Nassi. Morris II is the AI worm conceptualized in homage to the notorious 1988 Morris worm. In a controlled test environment, the researchers illustrated how their worm could hijack an AI email assistant to filch data and disperse spam. This break-in into generative AI systems like ChatGPT and Gemini unveils a significant chink in the digital armor of these tools. As Large Language Models (LLMs) evolve to process and produce not just textual but also visual and video content, the threat landscape diversifies. The creation of Morris II is an earnest warning to startups and tech giants alike that transformative AI systems are not impervious to exploitation. The novel methodology employed by generative AI worms involves 'adversarial self-replicating prompts.' These are inputs designed to compel the AI to generate output that includes additional instructions, that can then self-propagate the malicious task. Similar in its disruptive potential to SQL injections and buffer overflow attacks of yore, these worms can maneuver through AI systems by textual or multi-media means. The Cornell researchers' findings reveal that a seemingly innocent chatbot could be turned into a scammer, and safeguard measures within these systems are not yet foolproof against such inventive hack techniques.