CVE-2026-33017, a critical code injection flaw in the Langflow AI agent framework, was weaponized within hours of disclosure. CISA added it to KEV. Here's what OpenClaw users need to know about the accelerating AI supply-chain threat.
CERT/CC published four vulnerabilities in CrewAI — including a CVSS 9.6 critical RCE — that chain together through prompt injection. The flaws expose a systemic pattern: AI agent frameworks that silently downgrade security when infrastructure isn't perfect.
Between March 18 and 21, nine OpenClaw CVEs dropped — including a 9.9 critical that let any authenticated user become admin by asking nicely. A timeline, breakdown, and what it means for self-hosters.
Microsoft's March 2026 Patch Tuesday includes CVE-2026-26144, a zero-click Excel XSS that turns Copilot Agent into a data exfiltration channel, plus CVE-2026-21536 — the first critical vulnerability discovered entirely by an autonomous AI agent.
A high-severity vulnerability let any website silently connect to your local OpenClaw agent via WebSocket. Here's the technical breakdown and what you need to do.
A roundup of February 2026's OpenClaw security developments — critical CVEs, Microsoft's deployment guidance, the ClawBands oversight tool, and actionable hardening steps.