Leading AI lab security is inadequate to protect model weights

“Argued that leading labs' security was far from sufficient to protect model weights from state actors (paraphrase of "Situational Awareness: The Decade Ahead," June 2024).”