According to security experts, there is an ongoing cyber threat against agentic AI systems, whereby hackers are misusing the skill ecosystem of OpenClaw to spread malware. The use of third-party "skills" enables the adversary to compromise the system and turn the AI agent into an execution vehicle. This attack highlights the vulnerabilities posed by autonomous AI frameworks, which are fast evolving as new attack surfaces that lack proper monitoring. OpenClaw is an agentic AI system that operates based on a set of commands to perform tasks locally while automating processes by using large language models. It provides a skill system, which acts like plugins with defined instruction sets. They are installed using repositories such as ClawHub, giving them equal privileges to the AI agent. The threat actor uses the skill architecture of OpenClaw to infect their systems. The malware is embedded in the skill package to deliver infostealers, backdoors, and remote access capabilities. Threat actors disguise the malicious skills under the guise of legit software, including crypto mining tools and productivity apps, exploiting social engineering tactics.
Cybercriminals are using a signed Logitech installer to distribute a new Brazilian banking trojan known as TCLBANKER. The malware mainly targets users in Brazil and monitors dozens...
A newly disclosed Linux kernel vulnerability named “Dirty Frag” has raised serious security concerns due to its ability to enable local privilege escalation (LPE) attacks on af...
Threat actors have successfully executed a sophisticated prompt injection attack targeting AI-powered Web3 agents, resulting in the unauthorized transfer of nearly $200,000 in cryp...