AI, Cybersecurity, and Securing Model Weights with Miles Brundage and Chris Rohlf
In this episode, we’re joined by Miles Brundage, independent AI policy researcher and former Head of Policy Research at OpenAI, and Chris Rohlf, Security Engineer at Meta and cybersecurity expert. We cover the fundamentals of cybersecurity today (9:20), whether AI is tipping the offense-defense balance (21:00), the critical challenge of securing AI model weights (34:55), the debate over “AI security doomerism” (1:03:15), and how policymakers can strengthen incentives to secure AI systems (1:08:46).
--------
1:17:13
AI Safety Institute Rebrand, Congressional Hearing on Export Controls, and Meta's New Superintelligence Lab
In this episode, we discuss the U.S. AI Safety Institute's rebrand to the Center for AI Standards and Innovation (00:37), BIS Undersecretary Jeffrey Kessler's testimony on semiconductor export controls (10:36), and Meta's new AI superintelligence lab and accompanying $15 billion investment in Scale AI (22:26).
--------
31:37
Shield AI’s Ryan Tseng on Building an Autonomous Future for the DOD
On June 9, 2025, the CSIS Wadhwani AI Center hosted Ryan Tseng, Co-Founder and President of Shield AI, a company building AI-powered software to enable autonomous capabilities for defense and national security.
Mr. Tseng leads strategic partnerships with defense and policy leaders across the United States and internationally. Under his leadership, Shield AI secured major contracts with the U.S. Special Operations Command, Air Force, Marine Corps, and Navy, while expanding internationally with offices opening in Ukraine and the UAE.
Watch the full event here.
--------
50:51
AI in the “Big Beautiful Bill” and Safety Concerns About Anthropic’s Newest Model
In this episode, we discuss House Republicans’ proposed moratorium on state and local AI laws (00:57), break down AI-related appropriations across the executive branch (18:54), and unpack the safety issues and safeguards of Anthropic's newest model, Claude Opus 4 (26:51).
Correction: In this episode, a quote was incorrectly attributed directly to Rep. Laurel Lee (R-Fla.). The statement—“Should the provision be stripped from the Senate reconciliation bill, some Republicans are eyeing separate legislation.”—was reported by The Hill as a paraphrase of Rep. Lee’s comments.
--------
33:25
China's AI Industrial Policy, New Controls on Huawei's Ascend Chips, and President Trump's Middle East Trip
In this episode, we discuss Princeton researcher Kyle Chan's op-ed in the New York Times on China's industrial policy for AI and advanced technologies (0:35), what the Bureau of Industry and Security's new controls on Huawei's Ascend chips mean for China's AI ecosystem (10:09), and our biggest takeaways from President Trump's visit to the Middle East (19:07).
CSIS’ Gregory C. Allen, Director of the Wadhwani Center for AI and Advanced Technologies, is joined by cohost H. Andrew Schwartz on a deep dive into the world of AI policy. Every two weeks, tune in for insightful discussions regarding AI policy regulation, innovation, national security, and geopolitics. The AI Policy Podcast is by the Wadhwani Center for AI and Advanced Technologies at CSIS, a bipartisan think-tank in Washington, D.C.
Ireland