Your AI might be the biggest insider threat you’ve ever deployed without even knowing it. As large language models (LLMs) become embedded in cloud-native apps and infrastructure, a new kind of risk is emerging: data leakage through inference. With the right prompt, an attacker can extract sensitive data, proprietary logic, or even credentials from your model’s responses bypassing traditional cloud and application security. In this talk, we’ll explore real-world examples of prompt injection, model inversion, and inference-time exfiltration attacks. You’ll learn why AI models running in Kubernetes, serverless, or SaaS environments introduce hidden exposure, and what to do about it. Whether you’re a security engineer, cloud architect, DevSecOps lead, or AI/ML practitioner, this session will equip you to recognize and defend against one of the fastest-evolving threats in modern cloud systems. Because in the age of AI, the model doesn’t need to be hacked to become the leak.
Meletius Igbokwe is a cybersecurity-focused Modern Workplace Engineer with a strong passion for protecting organizational data and identities in today’s cloud-driven world. With expertise spanning cloud security, infrastructure, and automation, he is dedicated to helping organizations adopt modern technologies while maintaining a strong security posture. A professional member of BCS, The Chartered Institute for IT, and a holder of a Master’s degree in Cloud and Network Security from the University of Bolton, Meletius is deeply invested in advancing cybersecurity through research, innovation, and real-world application. His interests include the intersection of AI and cybersecurity, digital forensics, identity security, and securing enterprise environments using industry best practices. Meletius has contributed to several cloud transformation initiatives, using DevOps methodologies to enhance secure deployment and operational efficiency. He prioritizes resilience, compliance, and scalability in every solution he delivers.
