Friday, March 20, 2026
Managing AI trust, risk, and security
DETECTGOVERN
AI: What are the challenges for safe integration?
The emergence of artificial intelligence (AI) is transforming the digital landscape of businesses, but this technological advance raises crucial questions regarding the management of trust, risk and security.
At the first edition of Lyon Cyber Expo, we had the pleasure of bringing together Stanislas Chesnais – CEO – Xpdeep, Yosra Barbier – Regional Information Security Manager – Allianz Partners and Secretary of CEFCYS, Jonathan Caille – Technical Manager for the Command & Control product – Cyber division – Sopra Steria And Philippe Wieczorek – Innovation Director – Minalogic. They shared with us their perspectives on the challenges of using AI safely in businesses.
Trust: an imperative to build
In a context where AI is integrated into decision-making processes, trust becomes paramount. Philippe Wieczorek, Innovation Director at Minalogic, stated that Trust cannot be decreed; it is built through the transparency of algorithms and the explainability of the decisions made by these systems. "
It is essential that users understand how the algorithms work. and on what basis decisions are made. This need for transparency is particularly important in areas such as cybersecurity, where misinterpretations can have serious consequences.
Identify the risks to better mitigate them
The risks associated with the use of AI should not be overlooked. Stanislas Chesnais, CEO of Xpdeep, warned that “ AI can be a double-edged sword: it can be an asset for improving security, but it can also be a target for cyberattacks when its vulnerabilities are not identified. "
To address this issue, organizations must establish proactive risk assessment measures. This includes the implementation of detection systems capable of identifying potential vulnerabilities, as well as the adoption of a cybersecurity culture based on constant vigilance.
Data governance: a pillar of security
This sense of trust also rests on the quality of the data used to feed the AI models. Yosra Barbier, Regional Head of Information Security at Allianz Partners, emphasized that “ We must ensure that the data feeding our AI systems is not biased and reflects a diversity that will allow us to draw validated conclusions. "
Indeed, AI models trained on unrepresentative data compromise the reliability of the results provided..
This involves establishing processes for validating and verifying data, thereby ensuring its quality and relevance.
Awareness-raising and training for operational staff
Finally, Jonathan Caille, Technical Manager at Sopra Steria, emphasized the need to train users on the challenges of AI.Training users on the specifics of AI is essential to ensuring their trust.“He said. Good training allows users to better understand the tools at their disposal and to interact with them in an informed manner, thus minimizing the risks of misuse.
A structured awareness and training program should include elements such as understanding the algorithms used, the methods of data collection and processing, and identifying potential biases in decisions.
Towards a secure integration of AI
AI has the potential to transform our working methods and increase productivity, but its integration must take place within a framework that ensures security and reliability. It is crucial to foster collaboration between the various stakeholders, from regulators to end users, in order to build a trusted ecosystem that guarantees not only security, but also the responsible use of these emerging technologies.