Dark Prompts and Malicious Agents: Offensive AI in Action
This session explores the offensive potential of AI in the hands of attackers. We’ll dive into dark AI models like WormGPT, demonstrate how LLMs can craft personalized phishing attacks, and use AI to obfuscate malicious scripts. We’ll also showcase how agentic AI could enable autonomous, metamorphic malware and discuss the risks of hijacked AI agents and MCP (Model Context Protocol) connections. Each topic will be illustrated with real-world examples, live demonstrations, and practical mitigation strategies. Join us for a hands-on look at the emerging threats at the intersection of AI and cyber security.
