Orange Cyberdefense will be able to use a variant of Mistral Large 4 optimized for penetration testing, vulnerability analysis, and incident response. This partnership is announced as the French provider Mistral AI has just released the weights of its Large 4 model, meaning the parameters that determine its operation.
- Orange Cyberdefense obtient un accès anticipé à une variante de Mistral Large 4 pour les tests d’intrusion et la réponse aux incidents.
- Mistral AI publie les poids de son modèle Large 4 au même moment.
- Les équipes de sécurité gardent le contrôle des actions critiques, grâce à la supervision humaine et à la traçabilité.
- Les modèles tourneront sur des infrastructures européennes, mais aucune date de lancement n’est annoncée.
Mistral AI is partnering with the Orange group subsidiary, which has become the first European partner for designing managed security services. As part of its Design Partner program, Orange Cyberdefense will benefit from early access to a specialized version, designed to address several operational challenges.
Initial work will focus on penetration testing, attack simulations aimed at assessing a system’s robustness, as well as technical audits. The goal is to help experts explore more attack scenarios and identify vulnerabilities in information systems. The model will also be evaluated during post-incident investigations, particularly to speed up the analysis of technical traces and reconstruct cybercriminals’ modus operandi.
For Hugues Foulon, CEO of Orange Cyberdefense, this evolution responds to the industrialization of threats. “As threats become faster, industrialized, and now automated, defense must scale up. Volume, speed, trust: these are the three challenges cybersecurity faces in the AI era,” he states.
Uses governed by security teams
However, using an AI model in security operations requires several precautions. Orange Cyberdefense announces several governance measures: access management, operation traceability, model security testing, and human supervision. Critical actions will remain the responsibility of experts, who will retain human control over sensitive processing.
These controls are to accompany the integration of AI into the subsidiary’s security services. According to the two companies, the models’ analytical capabilities can help correlate technical events and identify clues that require further investigation.
European infrastructures for processing sensitive data
Beyond uses, the partnership emphasizes infrastructure location. The two companies plan to rely on computing capacities located in Europe to run the models and process information used in cybersecurity operations. This is a significant issue for organizations handling confidential data, whether event logs, vulnerability information, or traces collected during an investigation.
The use of European infrastructures aims in particular to meet requirements for confidentiality, regulatory compliance, and data control. However, Orange Cyberdefense does not intend to limit itself solely to Mistral AI’s models: the subsidiary indicates it maintains an agnostic approach in choosing its technologies, based on its clients’ needs and the specific constraints of each environment. Systems based on Mistral Large 4 will be gradually integrated into certain operational services offered to Orange Cyberdefense clients. The two partners, however, do not specify a launch date or a detailed deployment schedule.
