Securing AI Systems Through LLM Red Teaming: A Practical Pillar of Modern AI Governance
About This Session
As large language models and other advanced AI systems transition from experimentation into mission-critical environments, traditional security and compliance controls are no longer sufficient to address their unique risk profile. AI systems introduce new attack surfaces, including prompt injection, data leakage, model manipulation, and unpredictable outputs, requiring organizations to rethink how security and governance are operationalized.
This session presents a structured, practice-driven approach to securing AI systems through large language model red teaming as a core component of AI governance. The session is organized into three progressive modules: (1) foundational concepts and threat modeling for AI systems, (2) hands-on red teaming techniques, and (3) governance integration and operationalization. Participants will be introduced to established frameworks such as the NIST AI Risk Management Framework (MAP, MEASURE, MANAGE) and the OWASP Top 10 for LLMs to guide risk identification and testing strategies.
The practical component highlights specific techniques and tools used in LLM red teaming, including adversarial prompt design, jailbreak testing, and output evaluation using tools such as GARAK and DeepEval. Attendees will learn how to map testing activities to real-world risks such as sensitive information disclosure, model hallucination, bias, and system misuse. The session also demonstrates how red teaming outputs can be systematically documented and integrated into AI risk registers, impact assessments, and governance artifacts.
By positioning LLM red teaming as both a technical security function and a governance control, this session provides a repeatable methodology for embedding continuous testing into the AI lifecycle. Participants will leave with a clear understanding of how to structure red teaming efforts, select appropriate tools, and translate findings into actionable governance decisions that support safe, ethical, and resilient AI deployment.
This session presents a structured, practice-driven approach to securing AI systems through large language model red teaming as a core component of AI governance. The session is organized into three progressive modules: (1) foundational concepts and threat modeling for AI systems, (2) hands-on red teaming techniques, and (3) governance integration and operationalization. Participants will be introduced to established frameworks such as the NIST AI Risk Management Framework (MAP, MEASURE, MANAGE) and the OWASP Top 10 for LLMs to guide risk identification and testing strategies.
The practical component highlights specific techniques and tools used in LLM red teaming, including adversarial prompt design, jailbreak testing, and output evaluation using tools such as GARAK and DeepEval. Attendees will learn how to map testing activities to real-world risks such as sensitive information disclosure, model hallucination, bias, and system misuse. The session also demonstrates how red teaming outputs can be systematically documented and integrated into AI risk registers, impact assessments, and governance artifacts.
By positioning LLM red teaming as both a technical security function and a governance control, this session provides a repeatable methodology for embedding continuous testing into the AI lifecycle. Participants will leave with a clear understanding of how to structure red teaming efforts, select appropriate tools, and translate findings into actionable governance decisions that support safe, ethical, and resilient AI deployment.
Speaker
Kellep Charles
Cybersecurity Department Chair - Capitol Technology University
Dr. Kellep Charles serves as the Cybersecurity Chair as well as the Director of Center for Cybersecurity Research and Analysis (CCRA) for Capitol Technology University’s Cybersecurity department.
Dr. Charles (@kellepcharles) completed his Doctorate in Cybersecurity at Capitol Technology University and worked as a government contractor as an information security analyst for over 25 years in the areas of incident response, computer forensics, security assessments, malware analysis, and security operations.
He is the creator and executive editor of SecurityOrb.com (@SecurityOrb), an information security & privacy knowledge-based website with the mission to share and raise awareness of the motives, tools, and tactics of the black-hat community.
Dr. Charles (@kellepcharles) completed his Doctorate in Cybersecurity at Capitol Technology University and worked as a government contractor as an information security analyst for over 25 years in the areas of incident response, computer forensics, security assessments, malware analysis, and security operations.
He is the creator and executive editor of SecurityOrb.com (@SecurityOrb), an information security & privacy knowledge-based website with the mission to share and raise awareness of the motives, tools, and tactics of the black-hat community.
