Growing the Community of AI Hackers with Generative Red Team Testing

Watch a DEF CON 31 conference talk exploring the organization and significance of the largest live AI hacking event held at the AI village. Learn about the unprecedented collaboration between major tech companies like Anthropic, Google, HuggingFace, Meta, NVIDIA, OpenAI, and Stability, who provided models for testing, while Scale AI developed the platform. Discover how this groundbreaking event, attended by observers from the White House, NIST, NSF, and the EU, aims to expand the community of experts capable of effectively evaluating Large Language Models beyond simple prompt injections and jailbreaks. Understand the fundamental differences between AI and traditional software security, particularly in generative AI systems where stochastic components and acceptable error margins require new approaches to bug hunting, reporting, and bounty programs. Gain valuable insights into the history of algorithmic and bias bounties leading up to this landmark event, and receive practical tips for participating in the contest.