How to Break AI Systems - Before Someone Else Does

Learn to identify and exploit vulnerabilities in AI systems through this 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. Witness live demonstrations of prompt injection attacks that bypass security filters, AI agent manipulation techniques that transform helpful assistants into data theft tools, and privacy violations that expose sensitive information. Discover practical testing methodologies for evaluating your own AI systems, understand which attack vectors pose the greatest risks to production environments, and implement effective defensive measures. Gain hands-on experience with AI red teaming through access to a practice platform featuring vulnerable AI applications for continued skill development beyond the presentation.