PostHeaderSection

Symbol

Global NavBar

PostContentSection

Global Footer

Blog Detail Page

Beirit Harvey

AI isn’t just another tool in the security stack anymore, it’s reshaping how defenders think, work, and architect their systems.

Cyber Defense in the Age of AI

In a recent AI Security Council roundtable, “How AI Is Redefining Cyber Defense,” TensorWave CISO Cassandra Mack joined Chris Kirschke (CISO at Tuskira and AI Security Council lead) and Antoinette Stevens (Principal Security Engineer at Ramp) to share their thoughts. 

Together, they dug into how AI is changing fundamentals like identity, detection and response, threat modeling, and the future of SIEM along with where security teams should actually invest as AI-driven threats evolve.

This blog recaps the conversation, pulls out the key takeaways, and closes with a standout quote from Cassandra on where the real line is between helpful automation and harmful complacency.









Key Takeaways

The discussion opens with a foundational question: how is AI improving security fundamentals like identity, segmentation, visibility, and traceability?

Antoinette frames modern detection and response as fundamentally a big data problem and AI as “a cheap data scientist” that can reason over massive volumes of signals to find anomalies and threats that humans miss. Cassandra adds that when security leaders stop fearing AI and start using it, it changes how they see their environment: new tools, new baselines, and new ways of looking at their stack holistically instead of just fighting fires.

But both are clear: over-reliance on AI is a real risk.

If teams blindly trust model output, especially without domain expertise, they invite misinformation, bad decisions, and vulnerable code into production. AI lowers the barrier to build, but it doesn’t replace engineering or security judgment.

That leads into the core theme of the session: “human in the loop” is not a buzzword, it’s a safety boundary.

 * Antoinette draws a hard line at destructive actions: anything irreversible (like dropping a database or changing critical access rules) must require human approval.
 * Cassandra argues that AI should be treated like an nth-party in your supply chain and you don’t fully understand its internals, so you don’t give it unlimited authority. You define what it can and can’t do, you create escape hatches, and you validate its output just like you’d validate a managed security provider’s actions.

Identity and access control emerge as the next frontier:

 * There will be a clear split between permissions for humans and permissions for AI agents.
 * Even if a human has broad access, an agent acting on their behalf should have tightly scoped, auditable sub-permissions.
 * An AI agent’s impact is ultimately bounded by what identity it’s running as and what data it can touch.

The panel also touches on the future of logging and SIM in an AI-first world. As AI improves preventative security and governance, there’s a serious conversation about how much traditional SIM is still required versus what’s mandated purely for compliance. At the same time, Cassandra points out that cost pressures around data storage and processing can lead organizations to drop logs they later wish they had, especially when training and improving AI-driven defenses.

On where to invest, the panel keeps coming back to business alignment and fundamentals:

 * Cassandra advises starting from business strategy, not a generic “AI security” wishlist. In an AI compute company like TensorWave, that means investments that help win and protect customers looking to move off hyperscalers and onto specialized infrastructure.
 * At her previous SaaS startup, the focus was on experimenting with new AI security tools and challengers that could materially move the needle instead of just sticking with “safe” legacy vendors.
 * Antoinette emphasizes endpoint security and data protection, especially as engineers are increasingly willing to download tools or plugins simply because they claim to use AI.

Finally, the panel closes on how to keep threat models current when AI-driven attacks evolve so quickly. Antoinette’s answer is simple but not easy: constant learning. Reading, research, podcasts, and filtering the noise through what’s actually probable in her environment, not just theoretically possible. Cassandra highlights the importance of strong practitioner networks; CISO communities, councils, breakfast clubs, and local security groups where people openly share what they’re seeing and what actually works.

You can read the full white paper this conversation was based on here: AI and the Future of Cyber Defense


About TensorWave

TensorWave is the AI neocloud purpose-built for performance. Powered exclusively by AMD Instinct™ Series GPUs, we deliver high-bandwidth, memory optimized infrastructure that scales with your most demanding models–training or inference. 

Ready to get started? Connect with a sales engineer. 

Cyber Defense in the Age of AI

Product

Solutions

Resources

Company

Product

Resources

Solutions

Company