TLDR InfoSec · QUICK LINKS
OPENCLAW VULNERABILITY ALLOWED WEBSITES TO HIJACK AI AGENTS
Summary from the TLDR InfoSec newsletter
A serious vulnerability in the OpenClaw AI assistant enabled attackers to take control of agents through malicious websites by exploiting unsecured localhost WebSocket connections and unlimited password brute-force attempts, but it has been fixed in version 2026.2.25.