Security You Should Know

In this episode, Joseph Perla, founder and CEO at TrustedRouter, explains how his company's confidential-computing-based LLM router lets teams send prompts to hundreds of AI models while cryptographically verifying, rather than simply trusting, that no one, including TrustedRouter itself, can see the data passing through. Joining him are TC Niedzialkowski, former head of IT & security at Opendoor, and Keith McCartney, svp of security and IT at DNAnexus.

Want to know:

Why hasn't the AI privacy problem already been solved, given how many tools and architectures already exist?
What's wrong with the "zero data retention" promises most AI providers offer today?
How does a confidential virtual machine let a company verify, instead of just trust, that its own router isn't looking at its prompts?
For a use case like HIPAA-covered medical data, does TrustedRouter provide one complete, auditable path from a user's machine to the model and back?
How does remote attestation confirm end-to-end encryption down to the chip level?
Does TrustedRouter's SOC 2 Type 2 attestation cover the AI workload itself, or just its own SaaS interface?
Can customers restrict which data centers or jurisdictions their AI workloads run in?
What happens to a company's existing AI usage monitoring once it starts routing through TrustedRouter?

Check out the episode for the answers you need.

Huge thanks to our sponsor, TrustedRouter
TrustedRouter is the open-source AI gateway for teams that need privacy with proof. It routes hundreds of models through attested confidential compute, keeps prompts and outputs out of logs, verifies the code handling each request, and adds provider failover, spend controls, and OpenAI-compatible APIs for production AI systems.

In this episode, Dean Shroll, senior director of sales engineering at ThreatDown, explains how the company's Nebula and OneView consoles simplify protection across Mac, Linux, and Windows endpoints while its managed detection and response team absorbs the 24/7 monitoring that resource-constrained teams can't staff on their own. Joining him are Jonathan Waldrop, CISO at Acoustic, and Arif Hameed, CISO at C&R Software.

Want to know:

Why is managing security for resource-constrained organizations suddenly getting more attention?
How does ThreatDown keep its console simple without sacrificing depth across Windows, Mac, Linux, and cloud environments?
Where does ThreatDown draw the line between what it automates and what still needs a human analyst?
How does the platform distinguish suspicious behavior from normal DevOps activity without drowning teams in false positives?
What guardrails exist to stop overly broad exclusions from opening up a device?
Who owns what when an MDR provider is involved, and where does that responsibility actually end?
How should a CISO justify an MDR investment to the board when nothing bad happens?

Check out the episode for the answers you need.
Huge thanks to our sponsor, ThreatDown
ThreatDown delivers elite Managed Detection and Response purpose-built for resource-constrained teams with high-efficacy protection without complexity. Combining artificial intelligence and expert analyst judgment, ThreatDown intercepts sophisticated attacks with speed and accuracy, scaling security effortlessly. Recognized by MRG Effitas, AVLab, and G2, ditch fragmented tools for fast, 24/7 protection without overhead.

In this episode, Leo Taddeo, CEO and President at AppGate, explains how the company's zero trust network access platform cloaks internet-facing infrastructure from discovery, uses direct-routed architecture instead of cloud-routed hairpinning to avoid the latency tax, and layers in continuous, context-aware policy checks that go beyond device posture and MFA. Joining him are Ross Young, co-host of CISO Tradecraft, and Derek Fisher, Director of the Cyber Defense and Information Assurance Program at Temple University.

Want to know:

Why does securing remote access remain one of the top attack vectors year after year?
How does AppGate's direct-routed architecture avoid the latency and throughput limits of cloud-routed competitors?
What happens to a user's access when the circumstances that justified it change, like a closed service ticket?
How does cloaking network infrastructure stop an adversary's reconnaissance before it starts?
What's the fallback when a controller hosted in a customer's own data center goes down?
How does AppGate defend against credential replay and man-in-the-middle attacks that bypass MFA entirely?
What does AppGate's per-user licensing model mean for organizations managing multiple devices and enclaves?

Check out the episode for the answers you need.

Huge thanks to our sponsor, AppGate

AppGate secures and protects an organization's most valuable assets with its high performance Zero Trust Network Access (ZTNA) solution. AppGate ZTNA is the only direct-routed Zero Trust solution built for peak performance, superior protection and seamless interoperability. AppGate safeguards enterprises and government agencies worldwide. Learn more at appgate.com.

In this episode, Sagar Batchu, CEO of Speakeasy, explains how the Speakeasy AI control plane lets organizations adopt AI everywhere and still prove it's governed — putting every agent, tool, and MCP server behind a single security layer that authenticates each action, enforces policy, and inspects every session for prompt injection and data exfiltration. Pressure-testing the approach are George Finney, CISO at The University of Texas System, and Nick Espinosa, host of The Deep Dive Radio Show.
Want to know:
Why is knowing what your AI agent can access still an unsolved problem, and how much of it traces back to the identity and data governance we never fixed for humans?
Should an AI agent ever be allowed to take an action no individual can fully understand, even when it's statistically more effective than the human alternative?
Where's the line between legitimate governance and surveillance, and how do you assure employees you're not building a panopticon?
Are there decisions and departments that should stay permanently in human hands, off-limits to agents entirely?
When an agent acts through a company's API and something goes wrong, who's actually responsible: the employee, the company, the model developer, the API provider, or the governance platform?
Could a control plane have flagged the week's biggest AI incident before it escalated?
Is adopting this really as simple as switching on enterprise settings and plugging in an API, whether you're an SMB or the Fortune 1?
What is "cognitive surrender," and why does Speakeasy's CEO think it's the one quality companies most need to protect?
Check out the episode for the answers you need.
A huge thanks to our sponsor, Speakeasy.
Speakeasy is the enterprise AI control plane. It governs every AI agent, tool, and MCP server from one place. Every connection is authenticated, every policy enforced, and every agentic action inspected and logged. So organizations can scale AI adoption with visibility and control.

Copy for below Banner: AI usage is outgrowing your enterprise controls. Speakeasy is the AI control plane that governs every agent, assistant, and MCP server from one layer. Each connection is authenticated, each policy enforced, and every action recorded. So you stay in control as AI adoption scales."

In this episode, Curtis Simpson, CSO at Gambit Security, explains how Gambit moves organizations past that guesswork by continuously mapping a company's minimally viable business capabilities and validating whether the infrastructure, backups, and recovery processes behind them can actually deliver within the timeframes the business requires. Joining him to pressure-test the approach are Howard Holton, former CEO of GigaOm, and Adam Palmer, CISO at First Hawaiian Bank.

Want to know:

Why do so many disaster recovery plans hold up on paper but fail the moment they're actually needed?
What's the real difference between having something backed up and proving you can recover it?
What is a "minimally viable company," and how much of mapping it is automated versus built on human input?
What happens when your infrastructure fails today, versus what Gambit's roadmap has planned for tomorrow?
What should actually land in front of your CEO or board to demonstrate recovery confidence?
Is there an organization too small, or too mature, to get value from this kind of resilience assessment?
Why does decades-old infrastructure like the mainframe often cause the longest, most damaging outages?
Why is now the moment for security to finally move from a "trust but trust" model to "trust but verify" on recoverability?

Check out the episode for the answers you need.

Huge thanks to our sponsor, Gambit Security

Gambit is the AI-native cyber resilience platform for enterprises that can't afford downtime. It continuously maps your live environment, backups, security tools, and infrastructure-as-code, then validates whether your entire stack can actually recover from disruption. Gambit gives security and infrastructure leaders proof of recoverability on demand, so the business keeps running. Balens maps, controls, and proves your recoverability across every layer of your stack - against infrastructure failures, human and AI errors, and cyber threats. One live view of your cloud, IaC, and backups that updates as your environment evolves. Agentless deployment in 15 minutes. Learn more at gambit.security