BSidesCharm 2024 - Securing generative AI: threats, old and new
As we move closer towards generative AI becoming widely adopted, it’s important to understand the security implications, how they differ from more traditional cybersecurity, and where we can apply existing approaches to new systems and applications. This talk introduces AI security concepts, with a focus on LLMs, to equip participants with an understanding of the security landscape surrounding AI.
Presenter: Adam Swanda
Adam Swanda is a threat researcher with over 10 years working in cybersecurity, largely focusing on tactical and strategic threat intelligence. Adam is currently working as an AI Security Researcher at Robust Intelligence. He recently released the open source project “Vigil”, a Python library implementing various LLM defense measures for prompt injection and jailbreak detection.