How to Break AI Systems - Before Someone Else Does

Learn to identify and exploit vulnerabilities in AI systems through this comprehensive conference talk that demonstrates real-world attack methods and defensive strategies. Explore why AI systems fundamentally differ from traditional software in their inability to distinguish between instructions and data, making them susceptible to unique security threats. Discover practical attack demonstrations including prompt injection techniques that bypass security filters, hidden prompts embedded in documents, AI agent goal manipulation, and privacy violations that expose sensitive data. Master the threat landscape beyond simple chatbot vulnerabilities, understanding how helpful AI assistants can be transformed into data theft tools and how AI agents can be compromised in production environments. Gain hands-on experience with practical testing methodologies for evaluating your own AI systems, learn to assess which attacks pose the greatest risks to your applications, and develop effective defensive measures that provide real protection. Access a dedicated AI red teaming practice platform featuring vulnerable AI applications to continue developing your AI security testing skills beyond the presentation.