Your organization may have hundreds of AI agents running right now that your security team doesn't know exist. Every single one is an identity. Every identity is an attack surface.

In this episode of The Audit, co-hosts Joshua Schmidt, Eric Brown, and Nick Mellem sit down with Madhav Nakar, security researcher on the Phantom Labs team at BeyondTrust, to break down one of the most underexplored threats in enterprise security today: untracked AI agents creating exploitable "ghost identities." Madhav just returned from RSA — where he noticed every booth had an AI angle and a bubble forming — and he's here to cut through the noise with hard-hitting research and practical guidance.

🔍 Key Topics Covered:

• How low-code platforms let non-technical users spawn unvetted AI agents — and why that's a goldmine for attackers
• Ghost identities: what happens when AI agents run on untracked, over-privileged system identities
• The AWS sandbox DNS exfiltration proof-of-concept from BSides (BeyondTrust research)
• Why siloed AWS, Azure, and Okta teams create hidden privilege escalation paths
• "AI vs. AI" — the emerging defender model where autonomous systems monitor each other
• Browser extension cross-contamination and prompt injection risk for enterprise Claude deployments
• The three conditions that make any AI agent dangerous: private data access + untrusted instructions + tool execution
• Madhav's framework: inventory → least privilege → visibility — the basics that still matter most

Bonus: Madhav shares how "spiritually red-teaming yourself" — facing fear, breaking false narratives, and building trust — maps directly to how security professionals should approach zero trust and identity management. Plus: Joshua, Eric, and Nick on conquering stage fright and what that has to do with cybersecurity culture.

Don't wait for a ghost identity to become a ghost incident. Subscribe for weekly cybersecurity insights from practitioners, researchers, and the people defending the frontlines.

#GhostIdentities, #AIAgentSecurity, #NonHumanIdentity, #ZeroTrust, #TheAuditPodcast

What if the tools protecting your organization were the ones compromising it? In this episode of The Audit, co-hosts Joshua Schmidt, Eric Brown, and Nick Mellem — joined by IT Audit Labs team member Samuel Cala live in the St. Paul studio — unpack a wave of cybersecurity stories that all converge on one unsettling theme: trust is being exploited at every layer of the stack.

From an Iranian-linked APT group targeting U.S. healthcare infrastructure, to a sophisticated GitHub Actions supply chain attack that backdoored an AI coding library used by thousands of developers — the crew breaks down exactly how threat actors are weaponizing the tools, platforms, and third-party services organizations depend on daily.

They also dive into a disturbing revelation about AI-powered audit certifications: one company allegedly fabricated compliance evidence to hand out ISO 27001 and SOC 2 certifications at a fraction of the cost — raising serious questions about what those credentials are actually worth.

In this episode:

• 🇮🇷 Iran's escalation from cyber espionage to active disruption — what signals to watch for
• 🔗 The GitHub Actions / LiteLLM supply chain attack explained step by step
• 🧾 How an AI certification firm allegedly faked audit evidence — and what it means for your vendor trust
• 📡 FCC bans on foreign-made routers and the gray market hardware problem hiding in plain sight
• 🤖 OpenAI kills Sora — what it signals about where AI is actually headed

Whether you're a CISO trying to defend against nation-state threats or a developer trusting open-source libraries, this episode delivers the context — and the hard questions — you need to stay ahead.

Don't wait until your organization is the next headline. IT leaders need to stay ahead of evolving threats, and this episode delivers critical insights to help protect your business. Like, share, and subscribe for more in-depth security discussions!

#cybersecurity #supplychainattack #infosec #threatintelligence #ISO27001 #SOC2 #githubsecurity #irancyberattack #aicybersecurity #itauditlabs

A $25 million wire transfer. A fake CFO. An entire executive team that didn't exist. This is what modern cybercrime looks like — and your firewall won't stop it.

In this episode of The Audit, co-hosts Joshua Schmidt, Eric Brown, and Nick Mellum sit down with James McDowell — forensic psychology expert, cybercrime researcher, and adjunct professor at American Military University — to explore the chilling intersection of AI, human psychology, and cybercrime. James introduces the concept of "cognitive surrender": the slow, dangerous transfer of our thinking to AI tools, and how threat actors are exploiting it at scale.

What You'll Learn:

• What "cognitive surrender" is and why it's cybercrime's greatest accelerant
• How a $25M deepfake scam bypassed every red flag a trained employee had
• The psychology behind System 1 vs. System 2 thinking — and why attackers time their strikes around your lunch break
• Why voice passwords and family code phrases are becoming critical security tools
• How FraudGPT and dark-web AI models are lowering the barrier for cybercriminals
• What James's wave theory reveals about how we trust — and how that trust gets exploited

📖 Guest: James McDowell Forensic psychologist, cybercrime researcher, and author of Forensic Psychology and the Human Side of Cybercrime. James teaches at American Military University and leads research at [Research Institute] focused on the psychology of cyber offenders and victims.

📚 Book available on Amazon and Routledge. Search: Forensic Psychology and the Human Side of Cybercrime

Don't wait until your organization is the next headline. IT leaders need to stay ahead of evolving threats, and this episode delivers the psychological intelligence to help protect your business. Like, share, and subscribe for more in-depth security discussions!

#cybersecurity #cybercrime #socialengineering #deepfake #AIthreats #infosec #phishing #cyberpsychology #ethicalhacking #CISO