Subvolts take OpenAI's board member details real AI risks and safety checks before model releases. May 7, 2026 - Developers
- Learn about the security vulnerabilities and safety considerations OpenAI addresses before releasing new models.
- Business
- Understand the risks and governance challenges associated with deploying advanced AI systems.
- Everyday users
- Gain insight into how AI safety is managed and what potential risks exist with powerful AI.
This interview provides an inside perspective from an OpenAI board member on critical AI safety and security issues.
OpenAI Board Member Zico Kolter on the Real Risks of Frontier AI
Audio streams from Matt Turck. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
OpenAI board member Zico Kolter discusses frontier AI risks on The MAD Podcast. He explains OpenAI's safety review process for major model releases, the limitations of scale in improving safety, and how vulnerabilities like jailbreaks and AI agents expand attack surfaces. The conversation covers AI safety versus security, the debate around AI pauses, and the future of AI governance, offering a practical look at current AI development.
This episode explains how OpenAI checks if its most powerful AI models are safe before releasing them. It covers why bigger AI doesn't automatically mean safer AI and how hackers find weaknesses, which is important for understanding AI's future impact.
This episode offers a direct look into OpenAI's internal safety processes from a board member's perspective. It's for anyone concerned about the real-world implications of advanced AI, including researchers, policymakers, and the public. Kolter's insights connect to ongoing debates at all frontier labs about responsible deployment and the challenges of ensuring AI systems are secure and aligned with human values, especially as models become more capable and autonomous.
- HostMatt Turck
- GuestZico KolterBoard Member, OpenAI
- Guest: Zico Kolter, OpenAI Board Member
- Show: The MAD Podcast
- Length: 77 minutes
- Topics Discussed: OpenAI's safety oversight, AI preparedness framework, AI risk categories, AI safety vs. security, jailbreaks, AI agents.
What actually happens before a frontier AI model gets released — and who decides whether it is safe enough? In this episode of The MAD Podcast, Matt Turck sits down with Zico Kolter — OpenAI board member, Head of the Machine Learning Department at Carnegie Mellon, and co-founder of Gray Swan — for a deep conversation on the real risks of frontier AI. They discuss how OpenAI’s safety oversight works before major model releases, why more powerful models do not automatically become safer, how jailbreaks and prompt injection expose real weaknesses in AI systems, why AI agents dramatically expand the attack surface, and where frontier AI is headed next. A clear, practical discussion on OpenAI, AI safety, AI security, AI agents, frontier models, red teaming, reinforcement learning, and the future of AI governance.
(00:00) Intro
(01:32) OpenAI board role and Safety & Security Committee
(03:53) How OpenAI reviews major model releases
(05:33) OpenAI’s preparedness framework explained
(09:46) Are frontier AI models getting safer?
(12:33) Why AI safety does not come from scale
(15:23) The four categories of AI risk
(19:38) Doomerism vs accelerationism in AI
(24:11) The six-month AI pause debate
(26:20) AI safety as a global effort
(28:04) How Zico Kolter got into machine learning
(31:05) OpenAI in the early days
(34:14) Why Carnegie Mellon became an AI powerhouse
(38:43) What Gray Swan does in AI security
(40:44) AI safety vs AI security
(43:15) The GCG jailbreak paper
(49:19) How AI labs responded to jailbreak research
(50:19) State-of-the-art AI defenses
(52:32) State-of-the-art AI attacks
(54:22) Why AI agents expand the attack surface
(58:39) Are AI agents ready for production?
(59:40) Mechanistic interpretability explained
(1:02:31) Will AI be safer in two years?
(1:03:46) Reinforcement learning and self-improving models
(1:08:09) Do post-transformer architectures matter?
(1:09:29) Best research directions in AI now
(1:11:00) Zico Kolter’s Intro to Modern AI course
(1:14:53) Why modern AI is simpler than people think
- jailbreaks
- Zico Kolter
- Methods used to bypass AI safety restrictions and make it perform unintended actions.
- prompt injection
- Zico Kolter
- Attacks where malicious input tricks an AI into performing unauthorized actions.
- AI agents
- Zico Kolter
- AI systems designed to perform tasks autonomously, which can increase potential risks.
- red teaming
- Zico Kolter
- Testing AI systems by simulating attacks to find vulnerabilities before release.
38 min
45 min
43 min
Questions people ask
- How does OpenAI review major model releases?
- OpenAI reviews major model releases through its safety oversight processes, including a preparedness framework, before they are made public.
- Are more powerful AI models inherently safer?
- No, the episode suggests that more powerful models do not automatically become safer, and AI safety does not solely come from increased scale.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Matt Turck.