Join Maurits Kaptein, Co-Founder and CTO of Kyvvu, for an authoritative, research-backed conversation on solving the single biggest barrier to enterprise agentic AI: runtime safety and deterministic control. As AI agents transition from read-only chatbots to autonomous system operators capable of reading sensitive databases, sending emails, and executing API calls, traditional safety models are failing. Standard model guardrails and prompt filters evaluate text inputs, but they cannot predict or verify what an agent actually executes once it starts taking sequential, multi-step actions. Drawing on his background as a full professor of Applied Causal Inference, author of AI Agents at Work, and a serial founder with a successful exit, Maurits breaks down how Kyvvu’s Agent Security Kernel (ASK) enforces organizational rules at the execution layer—giving banks, insurers, and governments the confidence to deploy autonomous agents safely.
🎯 Insights You’ll Learn:
The Illusion of Prompt Filters: Why relying on system prompts, guardrails, or model alignment cannot stop agents from executing unauthorized system calls once exposed to indirect injection or unexpected reasoning paths.
Action-Level Policy Enforcement: How checking every API payload, database modification, and external communication against hard organizational constraints before execution prevents rogue agent behavior.
What Regulated Industries Demand: The strict compliance, auditability, and deterministic boundary conditions that financial institutions, healthcare providers, and governments require before deploying agentic workflows into production.
Causal Inference in Agentic Systems: How mathematical models of causality and path dependence help predict agent behavior trajectories and isolate system vulnerabilities before they hit production.
From Academia to Serial Entrepreneurship: Key lessons from transitioning from a professorship at Eindhoven University of Technology to co-founding Scailable (acquired by Network Optix) and building Kyvvu.
🌍 Why This Matters:
The true economic value of AI lies in autonomy—giving agents write access to enterprise tools so they can perform complex tasks end-to-end. However, giving an unconstrained, non-deterministic system direct access to production databases or financial APIs creates catastrophic operational risk. Maurits Kaptein and the Kyvvu team are filling this critical security vacuum. By introducing an in-process Agent Security Kernel that validates actions dynamically, Kyvvu enables enterprises to capture the productivity explosion of agentic AI without surrendering operational control.
👤 Expert Background:
Co-Founder & CTO of Kyvvu, building the foundational runtime security layer for enterprise AI agent deployment.
Acclaimed Author & Researcher, co-author of AI Agents at Work, author of Statistics for Data Scientists (Springer), and PhD recipient cum laude from TU Eindhoven including two years at Stanford.
🎙️ Hosted by Gary Fowler, CEO of GSD Venture Studios — global AI entrepreneur, investor, and innovation leader spotlighting founders navigating the realities of building in the AI era.
💡 Perfect For:
Chief Information Security Officers (CISOs), enterprise AI architects, compliance directors, CTOs, fintech and insurtech leads, and deep-tech venture investors seeking runtime governance solutions for autonomous agents.
🚀 Timely Topic:
As enterprise adoption shifts rapidly from conversational assistants to fully autonomous agents, traditional software security and perimeter defenses are no longer sufficient. Maurits Kaptein delivers a rigorous, actionable framework for governing AI actions at runtime.
Subscribe for more global founder conversations from GSD Venture Studios: https://gsdvs.com
#AISafety #AIAgents #Kyvvu #MauritsKaptein #AgenticAI #RuntimeGovernance #Cybersecurity #GaryFowler #GSDVentureStudios
