Past Events
Open-Source Assessments of AI Capabilities: The Proliferation of AI Analysis Tools, Replicating Competitor Models, and the Zhousidun Dataset
This paper presents an open-source methodology for analyzing military AI models, using the Zhousidun dataset to demonstrate how experts can replicate and evaluate AI capabilities.
New Tools Are Needed to Address the Risks Posed by AI-Military Integration
An exploration of confidence-building measures for military AI, and how these are complicated by divergent interpretations of risk and the complex roles of worldwide partners.
Taiwan Can’t Shake Its Nuclear Ghosts
Taiwan’s anti-nuclear political legacy has left it highly dependent on imported fuels and vulnerable to a potential Chinese blockade, forcing a debate over whether to revive a nuclear power program with a complex and fraught history.
Russian Nuclear ASAT Weapons: The Fallout
Reports that Russia is developing a nuclear-tipped anti-satellite weapon raised concerns about potential violation of the 1967 Outer Space Treaty, and promted a debate over whether existing international frameworks are sufficient in managing the growing risks of space militarization.
2/26/24 | NPWG hosts Dr. Erik Brubaker
Date: February 26 Location: UC Berkeley Risk and Security Lab, 2nd Floor Conference Room, 1995 University Ave. Time: 5:30-7 pm PT NPWG is thrilled to host Dr. Erik Brubaker from Sandia National Laboratories for a conversation about arms control verification...
Introducing the Risk Calculus
The Risk Calculus is a new podcast series from the UC Berkeley Risk and Security Lab. In Season One, join Professor Andrew Reddie for a deep dive into wargaming, an old way of thinking about risks that is being applied in new ways to some of today's most pressing...