Defenders strike back: AI-native security is closing the gap on adversaries

As AI accelerates enterprise transformation, it is simultaneously widening the attack surface organizations must defend — and compressing the time defenders have to respond.

The convergence of geopolitical tension, AI-powered adversaries and a surging wave of agent deployments has pushed agentic defense to the center of enterprise security strategy. The old playbook of manual detection and human-paced remediation is now nothing short of structurally obsolete, according to Morgan Adamski (pictured, right), U.S. cyber, data and tech risk leader at PricewaterhouseCoopers LLP.

“The adversaries may have the advantage, but I’m team defense,” Adamski said. “How do we integrate AI from the beginning to improve everything you’re doing from a security perspective so that you can benefit from the efficiencies and find the adversary faster?”

Adamski and Charles Carmakal (left), chief technology officer of Mandiant Inc., a subsidiary of Google Cloud, spoke with theCUBE’s John Furrier and co-host Alison Kosik at Google Cloud Next, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed agentic defense strategies, the evolving public-private cybersecurity partnership and why defenders now hold a structural advantage in the AI arms race. (* Disclosure below.)

Agentic defense and the shadow AI governance gap

The arrival of AI agents in enterprise workflows has introduced a new class of identity and behavior risk that security teams are not yet equipped to manage at scale. As PwC’s expanded alliance with Google Cloud signals, the industry is moving quickly to close that gap by embedding AI-native controls from the start of deployment. The core challenge is that of “shadow AI” — employees building and deploying agents outside of formal governance structures, according to Adamski. Organizations want to encourage their workforces to experiment with AI but lack the enforcement layer that makes that safe.

“Policies are supposed to be in place to help you move faster, not slower,” Adamski said. “You just have to write them the right way … write it to encourage innovation, but also to maintain that security barrier that you need.”

But the vulnerability exposure problem compounds the challenge. According to the Mandiant M-Trends 2026 report, adversaries have collapsed the defender response window from hours to as little as 22 seconds — a compression that makes human-only security untenable. The race is now between defenders using AI models to find and patch vulnerabilities before adversaries can exploit them, according to Carmakal.

“With AI, you can scale tremendously, well beyond what humans are able to do,” Carmakal said. “We’re going to find a lot of security gains by leveraging AI, but with that said, there’s always going to be risks associated with it too.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Google Cloud Next:

(* Disclosure: PwC sponsored this segment of theCUBE. Neither PwC nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE