Red-Teaming after Mythos — Zico Kolter & Matt Fredrikson, Gray Swan

This episode likely discusses red-teaming methodologies and adversarial testing approaches in AI systems, featuring security researchers Zico Kolter and Matt Fredrikson from Carnegie Mellon University's Gray Swan initiative. The conversation probably covers techniques for identifying vulnerabilities in AI models, the importance of adversarial robustness, and how organizations can proactively test their systems' safety and security before deployment.

Source: Latent Space | 2026-06-22