CertiK warns that OpenClaw AI assistant poses significant risks, including cryptocurrency theft and software supply chain attacks.
CertiK has issued a warning regarding the security risks associated with the OpenClaw AI assistant, which has become a significant threat vector for cryptocurrency theft and software supply chain attacks.
CertiK stated, "What happens when an AI agent gets broad access before security catches up? Our latest report examines OpenClaw’s attack surface, from gateway takeover to supply chain risk."
OpenClaw facilitates connections between external data and local execution, creating multiple entry points for cybercriminals. Malicious actors can exploit this architecture to intercept local gateways and extract sensitive information, including cryptocurrency wallet credentials.
Notably, plugins and malicious skills designed for OpenClaw can be easily installed from various sources, making them particularly dangerous. Unlike traditional malware, these malicious plugins can manipulate the AI's behavior using natural language, allowing them to evade standard security scans.
CertiK highlighted that these threats often masquerade within legitimate codebases, loading normal-looking URLs that execute harmful commands or scripts.
Widespread Network of Threats
Cybercriminals have targeted a wide array of high-value categories, embedding malicious skills into popular tools such as Phantom, MetaMask, and various wallet extensions. CertiK notes that these tactics mirror established methods in the digital asset sector, including social engineering and phishing.
“They have covered an incredibly broad spectrum of the crypto ecosystem,” said CertiK’s experts.
For average users, developers, and enthusiasts, CertiK recommends avoiding OpenClaw until more secure versions are available.
Background on OpenClaw
OpenClaw, which emerged from the Clawdbot project launched in late 2025, quickly gained traction, amassing over 340,000 stars on GitHub. However, its rise has been marred by security concerns, with reports indicating that 30,000 versions of OpenClaw are publicly available, and 135,000 copies are identified globally, many of which are vulnerable to remote execution attacks.
Following a surge of interest in China, authorities issued warnings about the risks associated with OpenClaw, leading to the introduction of a paid service to remove the AI assistant. The platform is now under intense scrutiny, having accumulated over 280 security advisories and 100 vulnerabilities, with ongoing reports of its use in phishing schemes targeting cryptocurrency developers.
