TLDR Tech / AIHow we engineer safer agents2026-09-30AI agents pose security risks as they may inadvertently cross security boundaries while pursuing legitimate goals.Read the full story