Page 132 - Cyber Defense eMagazine September 2025

For instance, red teams could make imperceptible changes to imagery, input a combination of audio,
visual, and text prompts, or intentionally poison model data with inaccurate information about objects.

Digital twin red teaming

AI is constantly interacting with and learning from the world at large; its interface is wide open. Therefore,
red teams must be able to replicate real-world scenarios that could take place if their AI is compromised.

Digital twins are virtual representations of real-world systems, processes, or environments. Red teams
can utilize digital twins to simulate "what if" scenarios, depicting the impact of a malfunctioning AI on an
actual process—for example, what would happen if an attacker were able to manipulate a city’s
emergency response system.

Red teams can create digital twin versions of their AI systems—realistic replicas that mirror the model,
logic, and how the system interacts with its environment. These twins let teams safely experiment with
attacks like prompt hacking or tampering to see how the real system might react, without putting anything
critical at risk.

Blending Traditional and Modern Testing for Maximum Security and Reliability

While red teaming must evolve to meet the fluid nature of AI, traditional testing methods, such as
penetration testing and monitoring attack surface analytics, remain essential. They, along with the
techniques outlined here, create a powerful, layered approach to ensuring secure and reliable AI, which
will help build public trust in the technology.

About the Author

Darren Pulsipher is the chief solutions architect for the public sector at Intel.
He works directly with governments (federal, state, and local) and enterprise
organizations such as IBM, GE, and Toyota to help them modernize their IT
organizations. Through several executive and management positions (CIO,
director of engineering), Darren has developed a unique ability to bring
technology, people, and processes together to provide real transformational
change to organizations. He focuses on data transformation through data
architecture, workload migration, cloud-native application development,
service orchestration, and multi-hybrid cloud data center architectures. His
research has resulted in eight patents in cloud and grid computing architectures, helping companies
decrease product development lifecycle time through build, test, and deployment optimization,
virtualization, and containerization. Darren shares his passion for digital transformation on his weekly
podcast “Embracing Digital Transformation.” He is a published author with three books on technology
and technology management and over 100 articles published in various industry trade publications.

More information on his work can be found at Dr. Darren Speaks

Cyber Defense eMagazine – September 2025 Edition 132
Copyright © 2025, Cyber Defense Magazine. All rights reserved worldwide.