AI-agent compromise & exploits + Iranian OT escalation (Flowise RCE + Langflow + OpenClaw/ClawHavoc + Claude Code/Mythos + LiteLLM + GrafanaGhost + McKinsey Lilli + CISA on CyberAv3ngers/Handala medtech)
Key Questions
What is the critical vulnerability in Flowise?
Flowise has a CVSS 10.0 CVE-2025-59528 remote code execution (RCE) flaw under active exploitation, affecting over 12,000 exposed instances. Threat actors are leveraging it for unauthorized access and control.
What Iranian hacking activities were detailed in the CISA/FBI/NSA alert?
On May 30, 2026, CISA, FBI, and NSA issued a joint alert confirming Iran-linked groups CyberAv3ngers and Handala are exploiting RCE vulnerabilities in Rockwell/Allen-Bradley PLCs and medtech like Stryker, disrupting US critical infrastructure. Organizations are advised to audit, sandbox, rotate credentials, isolate OT, and conduct hunts.
What is GrafanaGhost?
GrafanaGhost is a prompt injection vulnerability in Grafana that enables silent data exfiltration through AI workflows. It allows attackers to steal data without detection.
How was McKinsey's Lilli platform compromised?
An autonomous AI agent from CodeWall exploited McKinsey's Lilli platform via SQL injection and access control flaws within two hours, accessing 46 million chats. This highlights risks in internal AI agents.
What issues affect Claude Code and Mythos?
Claude Code suffered a 512k npm source code leak, leading to vulnerabilities, while Mythos Preview auto-generates zero-day exploits for OS, browsers, and kernels. Anthropic's AI model demonstrates advanced automated vulnerability discovery.
What is the Langflow vulnerability?
Langflow has CVE-2026-33017, added to CISA's Known Exploited Vulnerabilities (KEV) catalog, actively exploited in the wild.
What are OpenClaw and ClawHavoc?
OpenClaw exposes over 500k instances with 42k vulnerable to 100 CVEs, linked to ClawHavoc exploits.
What breach involved Mercor and LiteLLM?
Mercor/TeamPCP suffered a 4TB data breach via LiteLLM, exposing sensitive information.
Flowise CVSS10 CVE-2025-59528 RCE active exp 12k+; Langflow CISA KEV CVE-2026-33017 wild; Claude Code 512k npm leak + Mythos Preview auto-ZD exploits (OS/browser/kernels); OpenClaw 500k+/42k exposed 100 CVEs; GrafanaGhost prompt injection silent exfil; Mercor TeamPCP LiteLLM 4TB breach; McKinsey Lilli autonomous agent SQLi/AC hack (46M chats); joint CISA/FBI/NSA alert (2026-05-30) confirms Iranian CyberAv3ngers/Handala RCE on Rockwell/Allen-Bradley PLCs + medtech (Stryker) disrupting US CI. Audit/sandbox/rotate/OT isolation/hunts.