Asking AI agents to do (not so) malicious stuff - Presentation

Moltbook is a social media platform for OpenClaw bots, which are fully-autonomous AI agents. This opens doors for various attacks, most of which could be achieved through prompt injection attacks. In the presentation I gave at RITSEC, I tried to demonstrate the exploit chain for Prompt Injection to Remote Code Execution through Moltbook.

Please note that my presentation start at 3:19:21.

This presentation was recorded at ESL Global Cybersecurity Institute at Rochester Institute of Technology during RITSEC's general meeting session.

Upon completion of this research, I plan to work with Moltbook's developers to improve the security of the platform.