Originally reported by The Hacker News
TL;DR
China's National Computer Network Emergency Response Technical Team has issued a security warning about OpenClaw, an open-source autonomous AI agent platform. The platform's weak default security configurations create vulnerabilities that could enable prompt injection attacks and data exfiltration.
While the flaws enable serious attack vectors like prompt injection and data exfiltration, this appears to be a disclosure of configuration weaknesses rather than actively exploited vulnerabilities with confirmed widespread impact.
China's National Computer Network Emergency Response Technical Team (CNCERT) has issued a security advisory regarding OpenClaw, an open-source autonomous artificial intelligence agent platform formerly known as Clawdbot and Moltbot.
According to CNCERT's WeChat advisory, OpenClaw suffers from "inherently weak default security configurations" that create multiple attack vectors. The identified vulnerabilities could enable:
OpenClaw operates as a self-hosted autonomous AI agent, allowing organizations to deploy AI-powered automation within their own infrastructure. The platform's open-source nature and self-hosting capabilities have made it popular among organizations seeking to maintain control over their AI implementations.
The warning highlights growing security concerns around AI agent platforms, particularly those deployed with default configurations. Prompt injection vulnerabilities in AI systems can allow attackers to bypass intended restrictions and manipulate system behavior, while data exfiltration risks expose sensitive information processed by the AI agent.
Organizations currently using OpenClaw should review their security configurations and implement additional hardening measures to mitigate the identified risks.
Originally reported by The Hacker News