Thinking carefully before adopting agentic AI

Collection of NCSC advisories and blogposts (Apr–May 2026) highlighting four primary themes: 1) rapidly evolving AI risks and adoption — guidance to proceed cautiously with agentic and frontier AI, use oversight when using AI for vulnerability discovery, and build AI capabilities for defence; 2) urgent vulnerability and patching pressure — a call to prepare for a large ‘vulnerability patch wave’ and specific mitigation guidance for an unauthenticated RCE affecting F5 BIG‑IP APM; 3) covert/compromised edge infrastructure threats — multiple advisories on China‑linked covert networks and APT28‑ex