DiscoverSilver Bullet Security Podcast
Silver Bullet Security Podcast
Claim Ownership

Silver Bullet Security Podcast

Author: Gary McGraw

Subscribed: 1Played: 3
Share

Description

Welcome to the Silver Bullet Security Podcast, created and hosted by the Berryville Institute of Machine Learning, where we are defining the future of machine learning security.  From 2006-2018, we explored the nascent field of software security through the lens of building security in.  But today, the frontier has moved.  As we integrate machine learning into the fabric of our essential systems, we find ourselves facing a new set of architectural flaws and security challenges that traditional software security can’t touch.  On Silver Bullet, we’re shifting our focus to the security of machine learning—bringing the same deep-dive, “no silver bullet” philosophy to the world of AI.

https://www.garymcgraw.com/technology/silver-bullet-podcast/

Theme song is "Zagreb" by The Cheebacabra (https://linktr.ee/cheebacabra).

10 Episodes
Reverse
162 - Gregory Kurtzer

162 - Gregory Kurtzer

2026-11-0249:04

On Episode 162 of the Silver Bullet Security Podcast, BIML's Gary McGraw hosts Gregory Kutzer. Greg and I begin our discussion with the myth of "open source" AI models (open weights being just a small part of the solution). Open data are more important, and OpenWALDO gets this right. Recursive pollution is still a huge risk, and we talk about data provenance, cleanliness, and accountability. We then turn to harnesses and Agentic AI control. And we talk scientific computing and stochastic behavior.  In the end, we agree that good architecture requires human intuition, insight and inspiration. What will the black box do? * Gregory Kurtzer GitHub [https://gmkurtzer.github.io/] * OpenWALDO [https://openwaldo.org/] (weights, artifacts, licenses, data, origins) * CIQ [https://ciq.com/company/leadership/gregory-kurtzer] * CentOS in the press [https://www.theregister.com/os-platforms/2026/06/08/history-of-centos-how-a-biochemists-linux-hobby-project-became-the-enterprise-worlds-default-operating-system/5251530] * Rocky Linux Creator Built a New "Not Linux" Kernel in 11 Days Using AI [https://itsfoss.com/news/ginux/]
161 - Christoph Kern

161 - Christoph Kern

2026-10-0134:54

On Episode 161 of the Silver Bullet Security Podcast, BIML's Gary McGraw hosts Christoph Kern. Christoph talks about software security by construction and the importance of architecture, wiping out entire classes of bugs at Google, why mixing control and data in one untyped token stream is rather silly,  agentic AI security from tool wielders to massive swarms.  We end on a positive note, by emphasizing the important of human intuition, ingenuity and insight to architectural beauty. * Silver Bullet episode 100: IEEE Center for Secure Design [https://apothecaryshed.files.wordpress.com/2018/10/silverbullet-101-ieee-csd.mp3] * Christoph Kern - Google Research [https://research.google/people/christophkern/] * Secure by Design at Google [https://research.google/pubs/secure-by-design-at-google/] * IEEE Center for Secure Design Avoiding the Top Ten Software Security Design Flaws [http://bit.ly/ieee-CSD-gem]
160 - Aaron Bedra

160 - Aaron Bedra

2026-09-0430:26

On Episode 160 of the Silver Bullet Security Podcast, BIML's Gary McGraw hosts Aaron Bedra. Aaron talks about enterprise adoption of AI/ML from his perspective at the intersection of security engineering and executive leadership; the shocking pace of ML evolution; ML grinding on the security exploit front; MLsec (including recursive pollution); and the problem of Agentic swarm control.  MLsec meets the real world. * Aaron Bedra - DRW | LinkedIn [https://www.linkedin.com/in/aaronbedra] * Programming Clojure, Third Edition [https://pragprog.com/titles/shcloj3/programming-clojure-third-edition/] * abedra (Aaron Bedra) · GitHub [https://github.com/abedra] * BIML on recursive pullution [https://berryvilleiml.com/2026/01/10/recursive-pollution-and-model-collapse-are-not-the-same/] * Silver Bullet episode 097: Aaron Bedra [https://apothecaryshed.files.wordpress.com/2018/10/silverbullet-097-aaron-bedra.mp3], 4.30.14
159 - Melanie Mitchell

159 - Melanie Mitchell

2026-08-0127:31

On Episode 159 of the Silver Bullet Security Podcast, BIML's Gary McGraw hosts Melanie Mitchell [https://melaniemitchell.me/].  Melanie talks about the surprising progress made in AI since the '90s, whether real concepts emerge from data scale alone, whether modern connectionist systems extrapolate or only interpolate, and what humans do differently.  We also talk about the complex innards of modern transformer networks, measuring alien intelligences, role playing, sycophancy, and security, and why giving answers is easier than posing interesting questions. * FARG and Douglas Hofstadter [https://en.wikipedia.org/wiki/Fluid_Concepts_and_Creative_Analogies] * Copycat [https://mitpress.mit.edu/9780262132893/analogy-making-as-perception/] * The NeuriPS keynote [https://neurips.cc/virtual/2024/109414] * BIML's measurement work [https://berryvilleiml.com/docs/no-security-meter-ai.pdf] * Santa Fe Institute [https://www.santafe.edu/]
158 - Artem Dinaburg

158 - Artem Dinaburg

2026-07-0135:44

On Episode 158 of the Silver Bullet Security Podcast, BIML's Gary McGraw hosts Artem Dinaburg [https://www.dinaburg.org/about.html].  Artem talks about using Agentic AI to find, fix, and exploit software security defects.  We talk decompilation, stochasticism and the tension between harness development and LLM nondeterminism, human intuition, and the hard parts of program analysis.  We also talk about MLsec (machine learning security), recursive pollution, and tokenization.  Do we need more people in software security in the age of AI?  Yes we do. * Trail of Bits [https://blog.trailofbits.com/] * Mastadon [https://mastodon.social/@[email protected]] * Evaluating Solidity support in AI Coding Assistants [https://blog.trailofbits.com/2024/11/19/evaluating-solidity-support-in-ai-coding-assistants/] * Researchgate [https://www.researchgate.net/scientific-contributions/Artem-Dinaburg-2148442770]
loading
Comments