I promised the Board this AI was Secure: how to actually make it true
Thales webinar — online, April 30, 2026 (French edition 11:00 with Jonas Roblot; English edition 14:00 CEST with Stephen O’Connor)
It starts with a simple promise: the new AI initiatives deployed across the organisation are secure. The initial use case looks harmless — an AI assistant connected to the internal knowledge base. But as the organisation accelerates, the AI connects to business applications, retrieves sensitive data, integrates third-party services, and agents are introduced to automate tasks and trigger workflows. Guardrails are in place, security by design is applied, teams are trained — and yet, over time, the system becomes much harder to protect.
The webinar explored what happens when an AI reveals unauthorized critical data, when a malicious payload manipulates the model past its guardrails, when an AI system gains unsupervised privilege escalation, or when an AI assistant turns into a data exfiltrator. Through the story of a real enterprise AI deployment, we showed how the AI attack surface emerges during transformation projects and how to respond with practical controls: protecting the sensitive data used by LLMs, enforcing guardrails, deploying AI firewalls to filter malicious interactions, and monitoring AI usage across connected systems.
Sessions #
The webinar was delivered twice on April 30, 2026, one session per language:
- French session — 11:00, “J’ai promis au conseil d’administration que cette IA était sécurisée : comment m’en assurer ?”, co-presented with Jonas Roblot (Innovation and Strategy Manager EMEA) — Teams event page · LinkedIn announcement
- English session — 14:00 CEST, “I promised the Board this AI was Secure: How to actually make it true”, co-presented with Stephen O’Connor — LinkedIn announcement
Archived copy #
From the LinkedIn post by Thales Cybersecurity Products (archived July 2026):
It starts simple… an AI assistant helping employees work smarter. 🧠🛠️
But then it connects to apps, accesses sensitive data, triggers workflows and suddenly, the risk multiplies. ⚠️
Securing AI isn’t a one-time promise it’s an ongoing journey. 🔍
Join our upcoming webinar with Stephen O’Connor and Julien Ehrhart to explore how to protect data, enforce guardrails, and stay ahead of emerging #AI threats.
The French announcement ran with a short teaser video:
Campaign illustration of the risk the webinar tackles head-on — prompt injection against an enterprise copilot:
