Tag
Håkon Måløy discovered a prompt injection variant that turns into a self-replicating worm in Microsoft Word's Copilot, propagating hidden instructions across documents. Microsoft had 144 days to fix but no full mitigation.
University of Toronto researchers developed a proof-of-concept AI worm that uses a local open-weight LLM to autonomously reason about network vulnerabilities, generate tailored exploits, and replicate across hosts without human intervention, achieving 62% network infection in controlled tests.