Cyber X AI Reading Group

This week’s AI x Cyber reading group will discuss RAND’s 2024 Security Levels (SL) framework for protecting frontier AI model weights from theft and misuse. We’ll explore how the framework has aged, whether it has influenced real-world adoption, and whether its assumptions about attackers and defenses still hold. Discussion questions will include what the framework gets right, what it leaves out, and whether model weight security requires new approaches or can be addressed through existing cybersecurity principles.