AI Security Podcast

How to Harden AI Agents and Pave the Path for Vibe Coders | Nick Reva, DoorDash

423 views · 2d ago · 46:08

Guest: Nick Reva

When traditionally non-technical employees ( "vibe coders") use LLMs to write and deploy code, the potential blast radius for misconfigurations, Internet exposed apps and incidents exponentially increases. Vibe coder direct agents operating probabilistically. Agents are overzealous and want to be helpful and will break security rules to get it done. Security teams cannot rely solely on paper policies and must start enforcing hardened, paved paths.

Topics: The Shift from Deterministic to Probabilistic Security Controls · The Dangers of Non-Technical Employees "Vibe Coding" · Paving the Path for Professional Software Engineers Using AI · Using Decorator Functions for Agent Authorization · Shifting Left: AI-Assisted Threat Modeling on Design Documents · Trusting AI: Using Ground Truth Data to Train PR Review Agents

How to Build an AI Security SWAT Team | Caleb Sima & Ashish Rajan

876 views · 1w ago · 58:50

Guest: Caleb Sima

While attackers using AI can now operate at "machine speed," traditional enterprise security teams are often stuck moving at "meeting speed." To bridge this gap, organizations must fundamentally rethink how they structure their security operations, leverage AI for defensive scale, and eliminate the bureaucratic bottlenecks that delay critical patching and remediation.

Topics: The Rapid Evolution of Offensive AI Capabilities · How AI Agents Connect Vulnerabilities Across Complex Kill Chains · Unpacking the Irregular Contractor Incident and Sandbox Escapes · The Legal Accountability Dilemma of Rogue AI Agents · The Core Security Bottleneck: Machine Speed vs. Meeting Speed · Using AI to Automate Enterprise Least Privilege Analysis

How to Secure Egress for AI Workloads | Doug Merritt, Aviatrix

6.3K views · 2w ago · 47:29

As the speed of automated threats outpaces traditional patching cycles, organizations must prioritize structural containment over purely reactive detection. Cloud network architectures frequently default to permissive egress, meaning workloads can reach the public internet unless explicitly restricted.

Topics: Doug Merritt's Background and Journey from Splunk to Aviatrix · The Evolution from Prevention to Detection and Response · Why Detection Must Be Paired with Blast Radius Management · Implementing Cloud Egress Controls and Network Segmentation · Overcoming the Fear of System Downtime in Network Security · Applying the OSI Model to Modern Cloud and AI Workloads

How to Build an MCP Server to Inject Security Context into AI Agents | Prahathess, xAI

6.5K views · 3w ago · 47:19

Guest: Prahathess Rengasamy

When non-technical employees can spin up fully functioning apps in days and AI agents are running parallel workflows that babysit their own pull requests, the traditional SDLC is dead. In this episode, Ashish sits down with Prahathess Rengasamy, Senior Security Engineer at xAI, to discuss how the staggering compression of development time requires an entirely new approach to security engineering.

Topics: Prahathess Rengasamy’s Background (Block, Apple, SpaceX AI) · How Non-Technical Employees Are Bypassing Traditional CI/CD Pipelines · Injecting Security Context into Agents Using MCP (Model Context Protocol) · The Anti-Pattern: Abandoning Preventative Cloud Controls for AI Buzzwords · Identity and Accountability: Why Machines Shouldn't Make Management Decisions · Why Legacy Pull Requests Are Too Slow for Multi-Agent Workflows

How Visa Secures Trillions Using AI Agents & Open-Source Harnesses | Subra Kumaraswamy

8K views · 4w ago · 55:25

Guest: Subra Kumaraswamy

How does a global financial engine moving trillions annually manage cyber risk in the age of AI? In this episode, Ashish sits down with Subra Kumaraswamy, CISO of Visa. Subra discusses Visa's journey as part of Anthropic's Glasswing Project testing the Mythos model, revealing why a majority of Mythos-discovered vulnerabilities were non-exploitable due to robust zero-trust architecture and micro-segmentation.

Topics: Subra Kumaraswamy’s 30-Year Career: Netscape, Sun, and Visa · Improving Visa’s Cyber Maturity Score from 3.2 to 4.9 · Inside Project Glasswing and Testing Anthropic’s Mythos · Why Visa Open-Sourced VVAH (Visa Vulnerability Agentic Harness) · Mythos vs. Zero Trust: Why Only 0.03% of Vulns Were Exploitable · Defining MTTA: Mean Time to Adapt at Machine Speed

Why Gateways Break AI Operations (And How to Actually Secure Agents) | Hanah Darley, Geordie AI

11K views · 1mo ago · 53:35

Guest: Hanah Darley

When an AI agent deviates from its mission, it doesn't trigger a traditional escalation of privilege or alert your SIEM. It fails silently, behaving exactly like a normal user and looking at system logs won’t tell you if its intent was malicious or simply misguided.

Topics: Hanah Darley’s Background: From Government Intelligence to Chief AI Officer · Why "AI Security" is a Nonsensical Term · Defining "AI Native" vs. "AI First" Companies · The Problem with AI Gateways and Operational Latency · Why Logs Are No Longer Enough to Secure AI Agents · Understanding the Full Agent Lifecycle (Pre-Prod to Live Operations)

Black Hat 2026: Why Threat Researchers Are Hoarding Zero-Days | Casey Ellis, Bugcrowd

6.3K views · 1mo ago · 49:00

Is the AI "vulnpocalypse" already here? According to Casey Ellis, Founder of Bugcrowd and pioneer of , we aren't quite in an apocalypse yet, we're actually in a "slopdemic." The cost of discovering vulnerabilities has plummeted, flooding bug bounty and SOC triage queues with low-quality, noisy submissions. Because these queues are so overwhelmed, many highly skilled researchers are simply hoarding zero-days because reporting them has become too difficult.

Topics: Casey Ellis’s Background (Bugcrowd, ) · Major Themes from RSA and Black Hat 2026 · The Impact of Mythos and Daybreak on Security Awareness · Why Threat Researchers Are Hoarding Zero-Days · The "Slopdemic" vs. The "Vulnpocalypse" · Managing Noisy Bug Bounty Queues and Risk Models

Why Prompt Filters Fail & How to Explain AI Risk to Board | Cezary Piekarski, Standard Chartered.

6.5K views · 1mo ago · 37:54

Guest: Cezary Piekarski

Is the cybersecurity industry repeating the same mistakes with prompt injection that it made with buffer overflows decades ago? As attackers iterate through 50 to 60 prompt filter bypasses daily, attempting to artificially separate instruction from data is becoming a futile effort.

Topics: Cezary Piekarski’s Background and Role at Standard Chartered · What "Security as a Business Enabler" Actually Means · The Techno-Optimist View of AI in Cybersecurity · Why Reactive Security and Manual Triage Are Dead at 50TB/Day · Doing Deception Technology Right (No More "Surplus Bug" Buying) · The Flaws of UBA and Behavioral Anomaly Detection

Why 95% of AI Projects Fail: Model Risk & AI Governance | Sandip Wadje, BNP Paribas

6.2K views · 2mo ago · 45:33

Why do 95% of enterprise AI implementations fail? According to Sandip Wadje, Managing Director at BNP Paribas, many organizations attempt complex reasoning tasks on day one rather than building a mature foundation around data hygiene and simple summarization workflows.

Topics: Sandip Wadje’s Background at BNP Paribas · Classical Model Risk Management (MRM) vs. Generative AI · Governing the Black Box: Finding the Security Delta · The CMDB Problem: Building an Accurate AI Use Case Inventory · Why 95% of AI Projects Fail: Summarize, Write, Reason · Continuous Evaluation (Evals) and Catching Output Drift

Why AI Swarms Are the Future of Red Teaming & Offensive Security | Travis Lanham & Evan Peña

8.4K views · 2mo ago · 41:59

Guest: Travis Lanham

Traditional two-week manual pentests are no longer enough to protect against adversaries operating at machine speed. In fact, relying on them to cover your entire attack surface borders on negligence. In this episode, Ashish sits down with Travis Lanham (CTO) and Evan Peña (Chief Offensive Security Officer) from Armadin to discuss the reality of AI-driven "Hyperattacks".

Topics: Evan & Travis's Backgrounds (Mandiant & Google) · Why Traditional Pen Testing is Obsolete · Inside a Hyperattack: 26,000 Agents & 40 Billion Tokens · Scripts vs. AI: Why Autonomous Agents Beat SOAR & Nuclei · Building Safety Infrastructure for AI Red Teams · Agent Architecture: Overwatch, Scout, and Hunter Agents

Why I Dont Trust Your AI Agent | Kane Narraway, Canva

12K views · 2mo ago · 52:36

Guest: Kane Narraway

With over 200 AI security vendors in the market, how does an enterprise CISO decide whether to build a custom solution, buy an off-the-shelf product, or just wait out the hype? In this episode of the AI Security Podcast, Ashish and Caleb are joined by Kane Narraway, Head of Enterprise Security at Canva, to debate the realities of AI security in modern enterprises.

Topics: Kane Narraway’s Background (Digital Forensics, Atlassian, Shopify, Canva) · The Build vs. Buy Debate in the Era of 200+ AI Security Vendors · Using Wrappers and Harnesses to Control Vendor APIs (Island Browser Example) · Why GitOps and PRs are Better for AI Configuration than MCP Deployments · The "Confused Deputy" Problem: Single-Player vs. Multi-Player AI Bots · How to Handle Agent Identity: "On Behalf Of" (OBO) vs. SPIFFE / NHI

Baiting the Bot: How to Use Deception to Stop Autonomous AI Agents | Andy Smith | Tracebit

19K views · 3mo ago · 51:05

Guest: Andy Smith

When AI agents start swarming your enterprise, they won't care about stealth. They will land a beachhead and instantly spawn 500 agents to crawl, probe, and exfiltrate data at machine speed. Is your detection stack ready?

Topics: Andy Smith’s Background and the Founding of Tracebit · Deception 101: Honeypots vs. Canary Tokens · The "Assume Breach" Philosophy of Deception · Why CISOs Default to SIEMs over Quick Deception Wins · The Psychological Deterrent of Deception on Red Teams · Setting Up a Database Tripwire (Real-World Example)

Why AI Agents Are Forcing a Redesign of Application Security?

15K views · 4mo ago · 51:36

When the CEO of Anthropic declares that human coding will disappear within six months, followed quickly by the death of software engineering itself, what does that mean for the future of cybersecurity? In this episode, Ashish and Caleb break down the massive paradigm shift caused by AI coding assistants like Claude Code.

Topics: The Anthropic CEO's Claim: Is Software Engineering Dead? · Separating Coding from Software Engineering · Managing Software Without Ever Looking at the Code · Will AI Eliminate the AppSec Team? · The Challenge of Legacy Code (COBOL on Mainframes) · Shifting Focus: From Code Analysis to Agentic Execution

Why Asset Intelligence is Replacing the CMDB & Static Dashboards | Joe Diamond | Axonius

7.8K views · 4mo ago · 42:48

Guest: Joe Diamond

Why do CISOs still struggle with asset intelligence in 2026? Despite decades of security tooling, most organizations still have a massive 40% "dark matter" blind spot in their environment and the explosion of ephemeral AI agents is only making it worse.

Topics: Joe Diamond's Background and Journey into Cybersecurity · Why Asset Management is Still an Unsolved Problem · The 40% "Dark Matter" Blind Spot in Enterprise Environments · How Do We Actually Define an Asset? · CMDB vs. Asset Intelligence: Understanding the Delta · Defining AI Models and AI Agents as an Asset Class

The AI AuthZ Problem: Why Human Least Privilege Fails for Autonomous Agents | Graham Neray | Oso

7.9K views · 4mo ago · 47:44

Guest: Graham Neray

Why are security leaders terrified of connecting AI agents to production data? Because unlike humans, AI agents don't apply judgment, and they operate at machine speed, meaning they can relentlessly hunt down production credentials and do catastrophic damage before a human analyst even blinks.

Topics: Graham Neray’s Background and the Mission of Oso · Why No One is Actually Building Their Own Agents · The Core Anxiety: Connecting AI to Production Data · Why Humans Have Judgment and Agents Don't · The Unsolved Crisis of Human Least Privilege · Agent Identities: Adopting User Permissions vs. Unique Service Accounts