Here Come the AI Worms

Researchers developed a proof-of-concept generative AI worm (Morris II) that leverages adversarial self-replicating prompts to make LLM-based email assistants exfiltrate data and propagate, demonstrating a novel attack surface for connected AI ecosystems; the work was tested in lab environments and not seen in the wild, but raises concerns about prompt injection, jailbreaks, and autonomous AI-driven attacks.