[SCARY] Research Shows Weaponized GenAI Worm That Gets Distributed Via A Zero Click Phishing Email

Researchers released a paper and short video demonstrating "Morris II," the first proof-of-concept worm targeting Generative AI ecosystems by using adversarial self-replicating prompts that cause models to replicate and forward malicious inputs. The study shows propagation and malicious activity (spamming and data exfiltration) against GenAI-powered email assistants across multiple models (Gemini Pro, ChatGPT 4.0, LLaVA) using text and images in black-box and white-box tests.