AI agent poisoning, Langflow RCE exploited: Microsoft warns data theft risks

This briefing reports multiple high-severity, actively exploited issues across AI and enterprise tooling: AI agent tool-description poisoning enabling data exfiltration, Langflow unauthenticated RCE used for Monero cryptomining, GuardFall shell-injection bypasses against open-source coding agents, SimpleHelp authentication bypass delivering the Djinn infostealer targeting cloud/AI credentials, and Microsoft Defender (BlueHammer) zero-day exploitation in ransomware campaigns—urgent inventory, patching, sandboxing, credential rotations, and isolation are recommended.