A Showcase of Security Risks in GenAI-Powered System
Aliaksandr Matsiushonak, Lead Security Testing Engineer at EPAM, and Vladimir Fedotov, Security Architect at EPAM, gave a presentation at EPAM’s annual event — Learning Week 2023. Anywhere Club is now making the video* of their presentation available to registered members of the Club.
Large Language Models (LLMs) and other Generative AI products are exciting and show great promise. Including these components and services into applications can improve the interaction interface, by providing more human-like engagement, and can offer extended information processing capabilities. The integration of these technologies into business domains, however, requires heightened attention due to their complexity and the security risks they pose.
Vladimir’s comment on the presentation:
“GenAI is a fantastic technology. But if we adopt it without sufficient attention to security, we are doomed to experience more harm than value. Until now, humans have always been the weakest link in cybersecurity. Looking at the pace and direction of changes associated with GenAI, however, I think we may have found our successor. Unfortunately, there is no silver bullet. To address security issues, we need to apply the same layered and holistic approach that we do for traditional security. It is also vital to raise awareness about issues and about best practices at all levels. This talk is just one of the tiny steps in that journey.”
