Cyber Risks · The Hacker News

Open Claw Bug Enables One-Click Remote Code Execution via Malicious Link

Summary: OpenClaw, the open-source local AI assistant formerly called Clawdbot and Moltbot, patched a high-severity flaw tracked as CVE-2026-25253 (CVSS 8.8) in version 2026.1.29 released January 30, 2026; the Control UI trusts the gatewayUrl from the query string without validation and auto-connects on load, sending the stored gateway token to an attacker-controlled server. Researcher Mav Levin, credited with finding it, said a single malicious page triggers cross-site WebSocket hijacking because the server does not validate the WebSocket origin header, and the stolen token's operator.admin and operator.approvals scopes let an attacker turn off exec approvals, set tools.exec.host to gateway, and invoke commands on the host.

Bottom line: The exploit takes milliseconds after one click and works even when the gateway listens only on loopback, because the victim's browser makes the outbound connection - and the sandbox and approval guardrails users count on were built to contain prompt injection, not this, so they do not limit the blast radius.

Read the original at The Hacker News ↗

Topics
Vulnerabilities & Patching
Mentions
CVE-2026-25253
Related