Points clés à retenir
- Les stratégies de résilience traditionnelles s’effondrent face à l’ampleur,à la rapidité et à l’autonomie des systèmes basés sur l’IA.
- Les ransomwares industrialisés et les attaques basées sur l’IA ciblent désormais les systèmes de sauvegarde,sapant ainsi les fondements mêmes de la Recovery.
- Les organisations doivent passer d’une organisation en silos des équipes chargées de la sécurité et de la Recovery à un modèle unifié et continu appelé « opérations de résilience ».
- La résilience basée sur l’IA nécessite une visibilité des données en temps réel,une détection continue des menaces et une Recovery intelligente et propre à grande échelle.
- Les plateformes modernes permettent une Recovery rapide et vérifiée,aidant ainsi les entreprises à éviter le compromis entre une restauration rapide et une restauration sécurisée.
Alors que les entreprises se précipitent pour adopter l’IA,les RSSI et les DSI prennent conscience d’une réalité brutale : les stratégies de résilience qui fonctionnaient pour les infrastructures traditionnelles ne tiennent plus la route. Des systèmes qui pouvaient se remettre d’une attaque en quelques heures peuvent désormais mettre plusieurs jours. Les approches de sauvegarde conçues pour des données centralisées peuvent rencontrer des difficultés face à des charges de travail réparties entre les clouds et les plateformes d’IA. Parallèlement,les menaces et les vulnérabilités potentielles ne cessent de croître de jour en jour.Lors d’un récent webinaire,Tim Zonca,vice-président du marketing de portefeuille chez Commvault,a abordé une question urgente à laquelle sont confrontés les responsables de la sécurité : comment maintenir la résilience alors que l’IA bouleverse fondamentalement les règles du jeu ?
What Industrialized Ransomware and AI Mean for Resilience
CISOs and CIOs are under pressure. In spite of billions spent on cyber defense,nation-states and professional crime rings continue to reap ever larger payoffs from their victims. Ransomware-as-a-service has become widespread,and advanced AI automation is accelerating the industrialization of malware. By including backup systems in their attacks,adversaries are undermining the very foundation of resilience.
As attacks become more sophisticated,targets are becoming more vulnerable. AI is scaling faster than organizations can secure,with exponential data growth,fragmentation across environments,more complex supply chains,and autonomous systems operating with minimal oversight. AI agents and non-human identities now outnumber humans 80 to 1. When these systems make mistakes or expose vulnerabilities,the impact can cascade across interconnected business processes.
Breaches and failures are now almost inevitable; the only question is whether you can recover fast enough to keep your business running. For organizations using legacy systems that assume human-controlled systems,centralized data,and isolated failures,the answer may well be no.
Making Resilience Operational
As AI agents make decisions across the environment,including a significant number of errors,it’s no longer enough to focus on protecting infrastructure. Security leaders must now broaden their operational focus across three critical areas:
- Sécuriser en permanence les données à la source et surveiller les anomalies.
- Contrôler les identités des personnes,des entités non humaines et des appareils qui accèdent aux données et les utilisent de manière autonome.
- Assurer une Recovery prévisible des données à très grande échelle,sans compromission ni corruption.
Traditionnellement,la sécurité des données,la résilience des identités et la cyber-Recovery fonctionnaient comme des disciplines indépendantes,chacune disposant de sa propre équipe,de ses propres outils,de ses propres politiques et de ses propres exigences. Ces silos laissent des failles que les attaquants peuvent exploiter et ralentissent la récupération en cas de défaillance des systèmes d’IA. Pour combler ces lacunes,les organisations doivent unifier ces capacités au sein d’une boucle continue et automatisée. Nous appelons cette approcheles « opérations de résilience » (ResOps).
Les ResOps englobent trois exigences essentielles pour la résilience de l’IA :
- Understanding your data landscape: Knowing where data lives,its sensitivity,who’s accessing it (including AI agents and non-human identities),and what policies govern that access in real time. For AI workloads,this extends to protections like LLM prompt governance to control how models access data.
- Continuous threat detection: Automated systems that constantly monitor for anomalies,compromised identities,and data corruption. When AI systems are making thousands of autonomous decisions,you can’t wait for periodic security reviews.
- Intelligent recovery: Automated,comprehensive restoration for entire cloud-native applications and their dependencies. To prevent re-infection,teams must validate data integrity and conduct forensic analysis in an isolated cleanroom before moving trusted data back to production.
Enabling ResOps in practice
To help companies make the move to ResOps,Commvault has introducedCommvault Cloud Unity,la version la plus importante de notre plateforme à ce jour. Elle est conçue pour réunir les trois dimensions de la résilience :
- Réunirla sécurité des données,résilience des identités,and cyber-récupérationau sein d’un modèle opérationnel unique.
- Protecting all workloads,from both today’s production systems to tomorrow’s emerging AI stacks.
- Safeguarding data regardless of location,whether in clouds,regions,data centers,or edge locations.
A next-generation architecture brings AI automation to all facets of data protection,la sécurité des données,résilience des identités,and recovery. For security and IT teams,the platform provides simplicity at scale with one experience,one policy engine,and one interface designed to protect data,predict threats,and accelerate clean recoveries.
As security leaders know all too well,recovering from the most recent backup minimizes data loss but risks restoring compromised data. Rolling back to a verified clean state may eliminate threats but means losing hours or days of business-critical transactions or AI model training.
With Commvault Cloud,la surveillance continue des menaces and verified clean recovery points help eliminate this forced choice. The platform architecture automatically maps dependencies across distributed systems,helps maintain immutable backups,and helps enable one-click restoration of entire environments. Recovery can be both fast and clean,helping minimize loss as well as risk.
See ResOps in action
Regardez l’intégralité du webinaire à la demande to learn more about ResOps,explore the architecture and services of Commvault Cloud,and rethink your resilience strategy for the AI age.
FAQs
Q: What is Resilience Operations (Res Ops)?
A: ResOps is an operating model that unifies la sécurité des données,résilience des identités,and cyber-récupération into a continuous,automated discipline rather than treating them as separate IT functions. ResOps transforms resilience from a reactive response to incidents into an active practice that continuously understands data access patterns,helps detect threats and anomalies,and enables fast,intelligent recovery at scale.
Q: Why can’t traditional backup and recovery handle AI workloads?
A: Traditional backup tools were designed for centralized,human-controlled systems with isolated failures. AI workloads involve autonomous agents accessing distributed data across clouds and complex dependencies between microservices and containers,and they operate at a scale that manual processes can’t match.
When AI systems fail or are attacked,you need to recover not just data but entire application infrastructures with all their configurations,policies,and relationships – capabilities traditional backup tools lack.
Q: What does “unified resilience” mean in practice?
A: Unified resilience means bringing la sécurité des données,identity management,and cyber-récupération together under a single platform,policy engine,and operational model rather than managing them as separate functions with different teams and tools.
In practice,this provides a consistent approach to protect all workloads and data locations,automatically correlate security events with access patterns,and orchestrate comprehensive recovery that restores both data and the complete application infrastructure needed to use it.
Q: What’s the difference between cyber resilience and AI resilience?
A: Cyber resilience focuses on protecting infrastructure and recovering from security incidents,treating resilience as an operational state for confronting threats. AI resilience expands this to address challenges unique to AI-driven systems: autonomous agents making decisions with minimal oversight,exponential growth of data and non-human identities across environments,and cascading failures where problems in interconnected AI systems impact entire business operations rather than staying isolated.
Q: How does ransomware target backup systems?
A: Ransomware increasingly targets backup systems by exploiting compromised credentials with privileged access,moving laterally from production systems to connected backup repositories,or exploiting vulnerabilities in backup software itself. Modern ransomware families specifically hunt for backup infrastructure to encrypt or delete recovery points,preventing organizations from restoring clean data and maximizing pressure to pay ransom. This makes offline,immutable,or air-gapped backups essential for resilience.
Q: What is the clean vs. complete recovery dilemma?
A: The clean vs. complete recovery dilemma is the forced choice organizations face during incident response. You can recover from the most recent backup to minimize data loss but risk restoring compromised or corrupted data; or you can roll back to a verified clean state before the incident to eliminate threats but lose significant business-critical data. Traditional backup tools make organizations choose between completeness and safety,while modern resilience platforms aim to provide both simultaneously through la surveillance continue des menaces and verified recovery points.
Q: What is a cleanroom in cyber-récupération?
A: A cleanroom in cyber-récupération is an isolated,secure environment completely separated from production systems to help organizations safely test,validate,and analyze recovered data before restoring it to active use. Cleanrooms help enable forensic investigation of compromised systems,testing of recovery procedures,and verification that restored data is free from malware or corruption – all without risking reinfection of production environments or exposing sensitive data during analysis.
Sam Curcuruto is Director of Product Marketing at Commvault.
Blogs connexesRepenser la résilience à l’ère de l’IA
A CIO’s Perspective: Strengthening Business Resilience in the AI Era
