Joshua Saxe's Talk: Layered Defense for Securing Autonomous AI Agents

joshua_saxe · x · 2026-08-17

Joshua Saxe shares his keynote on securing AI agents that make dangerous decisions autonomously. He argues that alignment, sandboxing, and monitoring are necessary but insufficient; together, with gradual and controlled increases in autonomy, they could satisfice real-world safety.

Related event: Joshua Saxe: Layered Defenses Needed to Keep Autonomous AI Agents Safe(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →