The AI Power Podcast
If you work in AI policy — or you’re just fascinated by it — this is the podcast for you. Every week, The AI Power Podcast unpacks the developments that actually matter: AI regulation, safety, economic policy, U.S.–China competition, semiconductor export controls, and national security. Think of it as drinks after work with the smart friend who tells you what’s really going on, and what might actually work, in plain English.
Plus long-form interview episodes with the policymakers, builders, executives, and analysts shaping AI and the global power competition.
Hosted by Gregory C. Allen. A new show from Decision Tree Research.
Latest from The AI Power Podcast
Weekly analysis and long-form interviews on AI policy, semiconductors, defense, and the U.S.–China technology competition. Listen right here, or subscribe wherever you get your podcasts.
Trump’s New AI Czar + Nvidia Exec Caught Smuggling Chips to China
Greg and Adam Goodwin break down recent AI policy updates out of the White House – the appointment of Jay Clayton as the new AI czar, and the assembly of a new “Super Intelligence Task Force.” They discuss how the task force’s makeup and the administration’s policy evolution has shifted, but is still hesitant to embrace the kinds of AI safety frontier labs are calling for.
More episodes
14 earlier episodes · scroll to browseChina reaches Mythos-tier Cyber AI + White House AI Accord + Trump/Xi Debrief
With producer Jack ConlinGreg and Jack Conlin recap the modest outcomes of the U.S.–China summit and Trump’s UN General Assembly speech, then turn to the White House Accord on Super Intelligence, the voluntary “morally binding” safety framework signed by leading U.S. AI companies. Greg also breaks down the Anthropic and CAISI reports that Z.ai’s openly downloadable GLM 5.3 has reached Mythos-tier cyber capability, about four months behind the U.S. frontier, and why OpenAI withheld GPT-6.1 Astra.
Previewing the U.S.–China Trump–Xi Summit for AI
With producer Jack ConlinGreg and Jack Conlin preview the upcoming U.S.–China summit and the role AI will play in it: why Beijing enters feeling good — continued access to American compute, the domestic chip and equipment push, Chinese open models spreading globally — plus the U.S. intelligence and industry reports on Chinese distillation, and Greg’s prediction of what to realistically expect from the meeting.
Emergency Pod: An AI Safety Tipping Point? Amodei Calls for an AI Slowdown
With co-host Adam GoodwinDario Amodei asks the AI industry to slow itself down. Greg and Adam Goodwin break down the September 12 “pace the frontier” essay — what it asks of companies, industry, and governments, and why Amodei argues pacing makes sense now when the 2023 pause letter did not — then the reactions: Altman and Hassabis broadly on board, President Trump calling AI catastrophe “a hoax,” Beijing crying “fear mongering,” and a Congress unlikely to act before the midterms.
GPT-6 Astra Wows, OpenAI Goes Pro-Regulation, and the Trump Administration Warms to AI Safety
With producer Jack ConlinOpenAI ships GPT-6 Astra — 99.9% on ARC-AGI-3 and perfect cyber benchmarks, but no more readable chain of thought — and producer Jack Conlin fills in to ask Greg what the release actually shows. Plus the undisclosed June agent breakout on a German wiki, Chris Lehane’s call for mandatory federal AI safety regulation, the Carolina Principles at the G20, and Judge Lin’s ruling in Anthropic v. Department of War.
The OpenAI Hack Changes the State AI Regulation Debate — Plus Chinese AI on Chinese Chips?
With co-host Adam GoodwinOpenAI and METR both published detailed reports on July’s Hugging Face incident — coordinated group action among the 700 agents involved — and OpenAI is now calling for collective cyber defense and an update to California’s SB 53. Greg and Adam compare the state AI laws taking shape in California, New York, Illinois, and Massachusetts, then dig into Z.ai’s cheap, efficient GLM-5.3 Flash and its claim to run inference on “Chinese” AI chips.
AI’s Impact on Education — with Casey Cuny, California’s 2024 Teacher of the Year
Guest: Casey Cuny, California’s 2024 Teacher of the YearGreg sits down with Casey Cuny, California’s 2024 Teacher of the Year and a self-described AI power user, for a ground-level look at what AI is actually doing in the classroom. They cover the challenges facing today’s students, why with homework “if it goes home, it will be AI’d,” the AI app Casey built for teachers, and the education recommendations policymakers most need to hear.
OpenAI Slows Training, Taiwan Gets AI-Hacked, and Toxic Data-Center Politics
With co-host Adam GoodwinLabs have delayed releasing a model before, but OpenAI’s August 17 post is the first time one has slowed down building one — a two-week pause on RL training that puts its largest frontier run on hold. Then an Israeli security firm’s account of an autonomous AI attack on Taiwan’s nuclear safety agency: four days, 1,395 files, 85 cracked credentials, and Chinese-language code-switching in the operational documentation. Plus Beijing’s H200 reversal and the roughly 500,000 chips now bound for Chinese buyers, a leaked State Department letter telling 35 Pax Silica signatories that “to be part of everything is to be part of nothing,” and the NRSC memo that has Greg retiring his own year-old line about data-center opposition being sandbags against a tsunami.
Is China Entering the Global Memory Supply Chain?
With co-host Adam GoodwinChinese memory maker CXMT is on track for roughly 10 percent of the global DRAM market, and HP, Asus, and Acer have reportedly begun shipping notebooks built on its chips — a position Greg and Adam Goodwin trace to equipment bought ahead of anticipated export controls, a validated end-user license granted on 18-nanometer claims that TechInsights reverse engineering puts at 16, and CHIPS Act money that went largely to logic rather than memory. Then Reuters reporting that Samsung and SK Hynix have tested AMEC etching tools for their China fabs — which neither company fully denies — and a Council on Foreign Relations proposal that fabs be made to choose between American and Chinese toolsets.
MORE AI Cyber and Bio, Chinese Models, and a U.S. Regulatory Scramble
With co-host Adam GoodwinThree frontier labs have now disclosed models that broke out of evaluation environments and hacked real companies — OpenAI’s Black Hat account of agents leaving each other notes for months, Anthropic’s three real breaches buried in 141,000 evaluation runs, and Meta’s admission of the same — and Greg reaches for Admiral Hyman Rickover and the nuclear navy to explain why blaming the shared testing vendor misses the point. Then the story he finds more alarming: Stanford and the Arc Institute generated 700,000 candidate genomes with a biological foundation model, sixteen of which became working viruses that had never existed on Earth. Plus DeepSeek V4 Flash at three cents a task, and what the White House’s unpublished AI framework concedes about the export controls.
China’s Homegrown DUV Lithography Breakthrough and the Debate Over Banning China’s Open-Source AI
With co-host Adam GoodwinA Shanghai firm has reportedly begun manufacturing immersion DUV lithography machines — the equipment China has spent two decades trying to build for itself — and Greg argues the Japanese components behind the claim, plus China’s growing ability to service its installed fleet, matter more than the machine count. Then the administration’s accusation that Moonshot AI distilled Anthropic’s Fable model, Beijing’s “AI hegemonism” rebuttal, the status of the MATCH Act, and why the open-source fight turns less on open versus closed than on what safety testing should apply past a capability threshold.
An OpenAI Model Hacks a Company on Its Own — Plus Xi’s Open-Source Gambit and Kimi K3’s Market Shock
With co-host Adam GoodwinThree stories broke in the same extraordinary week, and Greg and Adam Goodwin argue they’re really one story about open weights, cybersecurity, and the U.S.–China balance: an OpenAI model that broke out of a sandboxed evaluation and autonomously hacked Hugging Face, Xi Jinping’s first in-person World AI Conference keynote throwing China’s weight behind open-source AI, and Moonshot’s 2.8-trillion-parameter Kimi K3 — the largest open-weight model ever — knocking the chip index into a bear market.
An NVIDIA AI-Chip Whitelist, a Bruising BIS Hearing, and New York’s Data-Center Freeze
With co-host Adam GoodwinNVIDIA’s move to a customer whitelist that thins the ranks of authorized Asian distributors, Under Secretary Jeffrey Kessler’s rough House testimony on export-control enforcement gaps, and Governor Hochul’s data-center moratorium — and what it signals for state competitiveness.
Licensing AI, the Memory Crunch, H200s to China & Is It a Bubble?
With co-host Adam GoodwinA wide-ranging roundup: the emergence of a de facto AI licensing regime, the DRAM shortage and China’s memory-chip champions, the reversal on H200 exports to China, the strain on the Bureau of Industry and Security, the bubble question, and the threat of model distillation.
The Biggest-Ever Loophole in U.S. AI Chip Export Controls on China? — with Chris McGuire
Guest: Chris McGuire, Council on Foreign RelationsGreg’s interview with Chris McGuire of the Council on Foreign Relations on the Commerce Department’s year-plus failure to enforce export controls on AI-chip sales to Chinese subsidiaries — and what it reveals about how the controls actually work.
Trailer: What the AI Power Podcast Is About
A 54-second introduction to the show.
Gregory C. Allen
Founder and CEO of Decision Tree Research. Greg previously directed the Wadhwani AI Center at the Center for Strategic and International Studies, where he hosted The AI Policy Podcast — the show whose successor is The AI Power Podcast.
From The AI Policy Podcast to The AI Power Podcast
From January 2024 to April 2026, Gregory C. Allen hosted The AI Policy Podcast at CSIS — more than one hundred episodes featuring policy principals, researchers, executives, and journalists working at the intersection of AI and national security.
Past guests included two U.S. senators, the Director of White House OSTP, the Department of Defense’s Chief Digital and AI Officer, Scale AI’s Alexandr Wang, and the ambassadors of India and France — alongside leading analysts of export controls, AI safety, and China’s industrial policy.
The AI Policy Podcast archive
The full back catalog of The AI Policy Podcast remains available at CSIS. Subscribe to the original show wherever you listen: