Latent Space · 2026-06-22 · notable
Latent Space: 'Red-Teaming after Mythos' — Gray Swan on AI security
Latent Space hosts Zico Kolter (OpenAI board, CMU) and Matt Fredrikson (Gray Swan CEO) to argue AI security is not 'cybersecurity with AI' — Gray Swan's Shade red-teaming model now beats human attackers at breaking frontier LLMs.
Latent Space episode on why AI security is its own discipline, with the Gray Swan team behind Shade and Cygnal.
What is it?
A 1h6m Latent Space podcast where Zico Kolter and Matt Fredrikson break down why securing AI models needs different tools than classical cybersecurity, drawing on Gray Swan's red-teaming products Shade, Cygnal, and the 15,000-attacker Gray Swan Arena.
How does it work?
Kolter and Fredrikson explain Simon Willison's 'lethal trifecta' — untrusted data, private context, exfiltration — and show that scaling models alone does not make them robust. Gray Swan's Shade automates adversarial attacks and now outperforms human red teamers; Cygnal sits in front of a model as a policy filter.
Why does it matter?
The episode lands as US export controls reshape who can deploy frontier models. Anyone shipping agents inherits the lethal trifecta by default, and the guests argue the first major prompt-injection breach is a 'gray swan' the industry can already see coming.
Who is it for?
AI security engineers, agent builders, applied researchers