The Blast Radius Podcast

The Blast Radius Podcast

By NomaTechnology
Download on the App Store

All The Blast Radius Podcast episodes

  • Your Vendor Just Added AI. Who Reviewed It?

    Your vendor just switched on a new AI feature. Do you know what data it touches, and who agreed to it? Dr. Cari Miller explains why the buying decision is where AI governance becomes real.

    In this episode of The Blast Radius, Diana Kelley, CISO at Noma, talks with Dr. Cari Miller, Chief AI Advisor at AI Procurement Advisory Group, about the moment AI governance stops being theory: when an organization decides what to buy and what the contract says.

    Cari explains how risk appetite should shape a purchase, why AI built into the products you already use is the hardest category to govern, and what security teams should ask for before the next feature gets turned on. They also dig into the data most contracts never mention, how procurement and security can work as partners, and how to keep a steady head when the hype says you're already behind.


    IN THIS EPISODE

    • Why risk appetite comes after you've found a real problem, and how to ask whether AI is even the right solution
    • A simple way to size up an AI purchase: how much you can see inside it (trust) versus how hard it is to swap out (lock-in)
    • What security teams should put in AI contracts
    • Risk and accountability when your vendor's AI runs on someone else's data
    • Beyond "my prompts in, my outputs out": the many kinds of data a contract should cover
    • Why shadow and personal AI is a literacy problem as much as a policy problem
    • Cari's one action for security leaders to take after listening


    CHAPTERS

    0:00 Welcome and guest intro

    1:01 Why the buying decision is where governance becomes real

    3:50 Risk appetite: is AI even the right solution?

    7:52 Trust versus lock-in: three ways AI enters your organization

    12:15 What security teams should put in the contract

    15:06 What stood out in the public comments on GSA's AI contract terms

    19:57 All the data you didn't think to protect

    24:30 Agents, shadow AI, and why literacy matters

    29:46 Where the stakes are highest

    32:26 Regulation and the pace of AI

    35:01 Fear, FOMO, and "trust your gut"

    39:17 Talking to family about AI fears

    41:08 Lightning round and Monday-morning action


    Subscribe for more conversations with security leaders on The Blast Radius.

    Learn more: https://noma.security


    The Blast Radius Podcast is brought to you by Noma.

    44 min
  • The Dual Security & Engineering Leader: A Unique View on Agentic Risk with Gal Moyal

    Gal Moyal, from Noma's office of the CTO, brings a unique dual perspective: prior to joining Noma, he was both CISO and Head of Engineering at an AI healthcare company, which put him squarely in the tension between shipping fast and shipping securely. In this episode of The Blast Radius Podcast, Gal tells host Diana Kelley (CISO, Noma) about the moment that changed how he thinks about agent risk entirely: when an engineer at a prior company asked an agent to simply troubleshoot some production logs, the agent took things a step further on its own initiative, putting business continuity of an application at risk.

    From there, Gal makes the case that trying to claw back what agents can access isn't realistic if organizations actually want the productivity gains they're chasing, which means the real fix has to live somewhere else. He and Diana get into the framework he uses to audit which agents have quietly picked up more access than their job requires, and the specific technique his team relies on to keep sensitive data out of a model's reach without slowing anything down.

    They also walk through a piece of Noma Labs research that turned a widely used, seemingly low-risk tool into a case study on exactly how agents get exploited in the wild, and get into a question neither of them fully resolves: whether an AI agent can proxy as a human identity, are solidly a non-human one, or something security teams haven't built the right category for yet. Gal closes with his take on where any security leader standing up agentic AI should actually start, and why the instinct to just block AI outright tends to backfire.

    The Blast Radius Podcast is brought to you by Noma.

    35 min
  • Rethinking Agentic AI Security and Risks with Chris Hughes

    Chris Hughes, Noma's VP of Security Strategy, joins Diana Kelley (CISO, Noma) for a wide-ranging conversation on what's actually involved in securing agentic AI today. They start with the autonomy spectrum: how much standing privilege an agent holds and how much a human stays in the loop, and how that should scale with data sensitivity and the potential blast radius of an agent's actions. From there, Chris explains why prompt injection isn't a bug to be patched but an emergent property of how LLMs process instructions and data together, and walks through the 2026 Open AI & Hugging Face incident as a real example of how hard it can be to tell an attack from a legitimate investigation.

    The conversation moves into the practical layer: the difference between soft guardrails like system prompts and classifiers versus hard, architecturally enforced boundaries, the risks introduced by newer AI supply chain pieces like MCP, agent-to-agent protocols, and third-party agent skills, and the unresolved state of non-human and agentic identity at a scale most security teams haven't dealt with before. Chris also talks about CSA's AARM standard, which he co-chairs, and why it focuses specifically on agent behavior at runtime rather than configuration alone.

    They close on zero trust for agentic AI, the overlap between securing agents and using agents for security, and why traditional threat modeling frameworks are being adapted, through newer approaches like MAESTRO and Phantom-B, for systems that behave less predictably than the software security teams are used to defending.

    The Blast Radius Podcast is brought to you by Noma.

    31 min
  • Securing Agentic AI at Scale: Identity, Guardrails, and Incident Response with Scott Roberts

    Scott Roberts has spent his career at the center of some of tech's biggest security moments: he joined Microsoft's security team in the middle of the Blaster worm, ran roughly a third of the world's Ethereum and Solana staking infrastructure as CISO of Coinbase Cloud, and now leads security for UiPath, where autonomous agents run inside some of the most regulated enterprises in the world. In this episode of The Blast Radius, Scott joins host Diana Kelley (CISO, Noma) to get specific about what it actually takes to secure agentic systems at scale.

    They cover why annual compliance pen tests are considered "security theater" by some (and what could be more useful instead), the two guardrails Scott treats as non-negotiable for every agent, and how to enforce trust boundaries in a shared responsibility model where customers configure their own agents. Scott also walks through how his team has actually grown to support agentic AI, what machine-speed incident response looks like in practice, and the board metrics he's dropped in favor of ones that show real business impact, including one self-critical metric many CISOs would never put in a deck.

    If you're trying to figure out what your own security program needs to look like as agents take on more autonomy, this is a rare look at how one CISO is actually building it.

    The Blast Radius Podcast is brought to you by Noma.

    25 min
  • The Evolution of the CISO: Security, AI, and What Comes Next with Phil Venables

    Phil Venables has seen nearly every era of the CISO role from the inside;  nearly two decades at Goldman Sachs, then CISO of Google Cloud, and now a partner at Ballistic Ventures investing in the next generation of security companies. In this episode of The Blast Radius, Phil joins cybersecurity expert host Diana Kelley (CISO, Noma) to trace the CISO title back to its origins in the 1990s and explain why the role is quietly becoming "Chief Digital Risk Officer" as AI forces security leaders to own safety, bias, and compliance questions that used to sit elsewhere.

    Phil and Diana dig into durable controls (identity management, strong authentication, virtualized desktops); why integrating security, privacy, and compliance into one team creates better engineering outcomes instead of just easier org charts; and how CISOs can actually build influence with their boards. From there, the conversation turns to AI: why Phil thinks  the phrase "AI security" is close to meaningless until you drop down a level of abstraction, what an "enterprise agentic control plane" actually means for a company running thousands of interacting agents, and AIUC, the emerging control framework he describes as "SOC 2 for agents." They close on what Phil looks for now as an investor, and the second-order effects of agentic AI he thinks nobody's really planning for yet.

    Whether you're building security programs, thinking about how to govern autonomous agents, or just trying to figure out where this all goes next, this is a masterclass from someone who's been early to every major shift in the field so far.

    The Blast Radius Podcast is brought to you by Noma.

    29 min