Builders and Breakers - A Collaborative Look at Securing LLM-Integrated Apps

Explore the dual perspectives of building and breaking LLM-integrated applications through this collaborative conference talk that combines defensive and offensive AI security expertise. Learn comprehensive strategies for securing Large Language Model applications as Rico Komenda, a senior security consultant, and Javan Rasokat, a senior application security specialist and penetration tester, demonstrate both attack and defense methodologies in real-time. Discover defensive techniques including prompt injection prevention, input validation frameworks, and continuous testing approaches to protect AI systems from adversarial attacks. Examine offensive perspectives through demonstrations of data poisoning, prompt manipulation, and generative misuse techniques that can lead to data leaks or unauthorized actions. Witness live demonstrations and analyze real-world case studies that showcase practical vulnerabilities and their corresponding security measures. Gain actionable insights into the evolving landscape of AI security, understanding how modern LLM integrations introduce unique security challenges while learning proven methods to mitigate risks. Whether you're developing AI-powered applications or conducting security assessments, acquire practical knowledge from two professionals with distinct but complementary expertise in application security, cloud security, offensive security, and AI red teaming.