Cybersecurity researchers at Check Point have demonstrated a novel technique where AI assistants like Microsoft Copilot and Grok can be leveraged as command-and-control (C2) relays. By using native components like Windows WebView2, malware can interact with these AI web interfaces to fetch instructions from attacker-controlled URLs. This method effectively turns legitimate, trusted AI services into proxies for malicious communication, making detection significantly more difficult for standard security tools.
This approach offers several advantages over traditional C2 infrastructure, including the lack of required API keys or accounts, which prevents defenders from easily revoking access. While AI platforms have built-in safeguards, researchers found that encrypting malicious data into high-entropy blobs can bypass these filters. Microsoft has acknowledged the research, recommending defense-in-depth strategies to prevent the initial infection that enables such post-compromise activity.
Top comments (0)