Does an AI rivalry between the US and China actually make us safer? Two AI policy researchers argue that Washington must pivot from power plays to partnership, working with Beijing to forge joint safety guardrails and avert catastrophic AI risks: ai-frontiers.org/articles/ameri…

x.com/i/article/1917…

Can the US Prevent AGI from Being Stolen? Securing AI weights from foreign adversaries would require a level of security never seen before. We explore what this would take. ai-frontiers.org/articles/can-t…

"Dynamism vs. stasis" is a clearer lens for AI safety debates. Helen Toner (Helen Toner) argues that many AI safety ideas lean too far toward control and rigidity—threatening a dynamic, open-ended future. Read it on AI Frontiers: ai-frontiers.org/articles/were-…

x.com/i/article/1922…

Interpretability research aims to reverse-engineer AI, yet despite a decade of effort, Dan Hendrycks argues progress has been minimal. Complex AI systems may simply defy a neat, neuron-by-neuron explanation, raising questions about the future of 'mechanistic interpretability.

Applications are open for the Summer (Jun 23–Sep 14) session of our free, online AI Safety, Ethics, and Society course! The course is open to all fields—no technical background required. Apply by May 30. More info & application: aisafetybook.com/virtual-course

Up to 50% of China's AI training is run on smuggled U.S. chips. Why? The US can't track chips after export. Congress just proposed a solution:  location verification. We break down how it could work: ai-frontiers.org/articles/locat…

x.com/i/article/1927…

x.com/i/article/1935…

x.com/i/article/1940…

x.com/i/article/1945…

x.com/i/article/1950…

x.com/i/article/1955…

x.com/i/article/1970…

Applications are open for the Winter (Nov 3–Feb 1) session of our free, online AI Safety, Ethics, and Society course! Open to all fields—no technical background required. Apply by October 10. More info & apply: aisafetybook.com/virtual-course

The term “AGI” is currently a vague, moving goalpost. To ground the discussion, we propose a comprehensive, testable definition of AGI. Using it, we can quantify progress: GPT-4 (2023) was 27% of the way to AGI. GPT-5 (2025) is 58%. Here’s how we define and measure it: 🧵

x.com/i/article/1978…