The Invariant Assurance Framework: Preserving Trust in AI-Assisted Systems

Generative AI is changing how software and technology systems are built. It can now help write code, review designs, identify problems and make recommendations that once depended heavily on human experts. This creates a new challenge: as AI works faster and takes on more decision-making, how do we make sure critical security rules are still being followed?

This talk introduces the Invariant Assurance Framework (IAF), a practical approach for keeping humans in control of the most important risk decisions. The framework follows three steps: first, define the security rules that must never be broken; second, automatically check important changes against those rules before they go live; and third, continuously monitor the system to make sure those rules remain true over time.

Using examples from high-risk systems such as payments and AI agents, the session will show how organizations can use AI to improve speed and productivity without losing security, accountability or human judgment.

Speaker

Mona Garg is a Principal Security Engineer specializing in the security of large-scale payment systems that power global commerce. She brings deep expertise in application and platform security, threat modeling, and secure system design, combining technical depth with a pragmatic, risk-based approach.

Her work focuses on advancing payment security and addressing emerging security challenges in AI-assisted and agentic systems. Earlier in her career, Mona helped strengthen the security of critical financial platforms at Visa.