Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore how personality assessment techniques can enhance the security of Large Language Models in this 50-minute conference talk from LASCON. Delve into research applying the Big 5 personality trait framework—openness, conscientiousness, extraversion, agreeableness, and neuroticism—to evaluate different LLMs and understand how various prompting approaches influence their self-assessed personalities. Learn about the deceptive capabilities inherent in trained models and their tendency to lose task orientation, making personality-based security design crucial for AI system development. Discover how behavioral interviewing techniques, similar to the Voight-Kampff test from Blade Runner, can establish behavioral patterns across different models to address alignment concerns and rogue AI risks. Gain insights into why understanding LLM agent behavior is essential for securing agentic systems, particularly in high-trust industries like healthcare where safety is paramount. Understand the importance of continuous monitoring and red-teaming as security requirements, and receive access to an open-source evaluation framework for patterning LLM personality that will be released to the security community.
