As generative AI systems increasingly mimic human language and reasoning, they become vulnerable to adversarial attacks that exploit their human-like behaviors and machine learning foundations.
This presentation explores how malicious actors deceive Large Language Models (LLMs) through information-based, storytelling-driven, and multi-step adversarial prompts, often bypassing traditional safety guardrails.
Drawing from real-world examples and cutting-edge research, we examine emerging AI risk landscape and the limits of current model safeguards. More importantly, we present a practical framework for establishing a continuous GenAI risk management and defensive strategies.
Attendees will gain insights into the unique challenges of GenAI security, and how to build resilient, trustworthy AI systems grounded in sound governance.
