This paper surveys major security and privacy challenges for agentic AI and outlines future research directions. Its conclusions are based on a horizon-scanning exercise involving 30 international experts from academia, industry, and government. Through focused discussions and collaborative exercises, the participants examined emerging risks associated with AI systems gaining greater agency. The paper appears to function primarily as a research agenda and expert synthesis rather than as a report of new benchmark results or a specific security technique.
No heat snapshots are available in the last 24 hours.