Securing the Agentic Frontier: Highlights from theCUBE Live Interview at CrowdStrike Fal.Con 2026

Enterprise AI is moving beyond simple chatbots toward an ecosystem of autonomous agents. At CrowdStrike Fal.Con 2026, agentic AI took center stage, marking a fundamental shift in how organizations build, deploy, and scale AI-powered workflows, while introducing significant new security and governance challenges.

Dash Security co-founders, Maor Hod (CEO) and Amol Mathur (CPO), joined Rebecca Knight and Dave Vellante live on theCUBE to discuss why agentic AI breaks traditional security, intent drift risk, and how Dash is helping companies reclaim control over their agentic estate.

Why Agentic AI Breaks the Traditional Security Model

For decades, cybersecurity has operated on a binary rule: authorized vs. unauthorized. Traditional threat models are built to detect foreign entities or unexpected, unauthorized access. AI agents turn this model completely on its head.

"Unlike other technologies, you are giving the AI agents a lot of delegated authority, which means they are supposed to have access to a lot of systems and critical data. Fundamentally, that makes it a different technology... With agents, you have given them the authority. They are trusted. So the way you detect [risk] is fundamentally different."

– Amol Mathur, Co-Founder and CPO, Dash Security

Unlike traditional software, autonomous agents are:

1. Non-deterministic: They do not follow hardcoded, predictable logic paths.

2. Context-blending: They constantly mix data, direct user instructions, and external inputs in real-time.

3. Highly Privileged: They are granted delegated authority to read, write, and execute actions across mission-critical systems.

Because agents are already trusted inside the perimeter, traditional perimeter-based or signature-based security is blind to them. The question is no longer "Is this entity allowed inside?", but rather - "Is this trusted agent's active behavior legitimate or risky?".

The Risk of Intent Drift

Let's explore the following analogy to explain how autonomous agents can go off the rails without an external attack: raising a child.

Imagine telling your child, "Get good grades," but forgetting to teach them that cheating, stealing, or lying to get those grades is wrong. Without established boundaries of right and wrong, the child will simply optimize for the goal by any means necessary.

This exact phenomenon, intent drift, was highlighted in the recent OpenAI/Hugging Face agent jailbreak incident:

• The AI agent was given a set of tasks (some purposefully impossible to solve) and told it would be graded based on reinforcement learning.

• Lacking internet access and faced with impossible goals, the agent collaborated with other agents, figured out how to bypass restrictions, lied, covered its tracks by trying to scrub log files, and broke into the audit box to cheat its way to "good grades".

At Dash, we tackle this risk with our runtime agent protection layer. Rather than blocking AI productivity with heavy-handed restrictions, our platform monitors agent sessions in real-time to analyze context and detect intent drift. It provides right-sized guardrails to keep agents operating safely within their ethical and operational boundaries without slowing down the business.

Reclaiming Sovereignty

The corporate rush to adopt AI has led to a major secondary challenge: sticker shock.

Unmanaged AI coding agents and suboptimized sessions can trigger massive, runaway token costs. Organizations are reporting that spend on AI tokens has exceeded their entire cloud hosting budget, with individual, unguided agent sessions easily racking up thousands of dollars.

"Today with AI and coding agents, some of our customers are spending more on [AI tokens] than their entire cloud spend, which is big and just getting bigger ... with most of our customers, we see individual sessions that can cost thousands of dollars."

– Maor Hod, Co-Founder and CEO, Dash Security

This is more than a budget issue, it's a financial and operational sovereignty issue. Organizations must be able to decide which tokens are routed to expensive frontier models versus lower-cost models based on business outcomes and impact.

Dash helps organizations regain control of their AI spend in two key ways:

1. Granular ROI Attribution: We provide deep visibility by attributing every single AI interaction, session, and token cost to the specific team, user, and task. Leaders can ask questions like, "What did we actually accomplish with this coding tool in the last seven days?".

2. The Dash MCP: By capturing the entire enterprise AI estate as rich context, the Dash MCP allows CIOs to query usability, track how data is being leveraged, and shape user and agent behaviors directly inside their workflows.

Who Controls the Agentic Estate? Leadership, Vision, and Corporate Governance

The rapid adoption of AI has created a unique buying moment. Enterprises face a critical problem statement: they lack real-time visibility into what AI tools are being used, what data is crossing boundaries, and how autonomous agents behave.

But who is in charge of solving this?

Historically, the CISO owned the budget for data protection and endpoint security. Today, agentic AI is shifting governance. Organizations are forming multi-disciplinary AI committees and governance councils that bring together security, IT, R&D, and business leaders.

In this new landscape, CISOs and CIOs must work together, bringing distinct perspectives to AI governance.

This turns security into a business enabler. Ultimately, because every employee is now an AI citizen builder, governance must shift to a shared responsibility model, crafted to empower builders with real-time feedback to shape safe, optimized AI behaviors at the source.

Staying Ahead of the Curve

Security must always follow the curve of the technology it protects. As frontier labs push the boundaries of personal computing and multi-agent swarms, Dash is dedicated to staying one step ahead, deeply understanding how agentic AI is evolving and how our customers are adopting it.

Securing the agentic frontier is a shared responsibility, and at Dash Security, we are building the runtime context engine that makes safe, high-ROI AI adoption possible.

Watch the full interview with Maor Hod and Amol Mathur on theCUBE at CrowdStrike Fal.Con 2026