For a long time, maintaining a system mainly meant keeping it available. Monitoring basic metrics, fixing incidents, applying patches when they arrived. As long as the service was responding, the system was considered maintained.
This view is no longer sufficient.
Today, a system can be technically “up” while being increasingly out of control. Dependencies age without anyone tracking them. Service accounts retain permissions granted “temporarily” three years ago. Logging covers application errors but not sensitive access. New regulatory requirements pile up without being absorbed. And every change becomes slightly riskier than the last, because knowledge of the system is concentrated in a few people.
Trust-condition maintenance starts precisely from this observation.
Availability is just one indicator among many
A good availability rate is necessary. It does not prove that a system is healthy.
Take a concrete example: a contract management system in the insurance sector. The SLA shows 99.7% availability over the past twelve months. The team is satisfied. But beneath the surface, Node.js dependencies have not been updated in eighteen months, including two libraries with published critical CVEs. Production database access rights are shared through a generic account used by four people. And the last backup restoration test dates back to the initial deployment, three years ago.
The system holds. But it is accumulating security debt, documentation debt, and operational debt. The day an audit, a serious incident, or a new regulatory requirement arises, the catch-up will be brutal.
The issue is therefore not just about “keeping things running.” It is about keeping them running sustainably in a changing context.
Operational maintenance is not trust-condition maintenance
The distinction between operational maintenance and trust-condition maintenance is not merely semantic.
Operational maintenance answers a simple question: is the system working? It covers monitoring, incident resolution, minor version upgrades, and routine operations. It is the essential foundation.
Trust-condition maintenance answers a broader question: is the system still under control? It includes operational maintenance, but adds:
- regular review of dependencies and their vulnerabilities;
- verification that access rights align with actual roles;
- control of log quality and their usability in case of audit;
- the ability to absorb a regulatory change without a rewrite;
- management of sensitive changes with a clear approval and rollback process.
The Log4Shell episode illustrated this gap strikingly. Teams operating in pure operational maintenance mode spent weeks identifying which systems used the vulnerable library. Those that had an up-to-date inventory of their dependencies and a proven update process were able to respond within days.
Maintaining also means absorbing change
A significant part of trust-condition maintenance consists of preserving the ability to evolve.
A system that no one dares to modify is not truly maintained. It is simply preserved in its current state until the next serious incident. You can recognize it by concrete signs: framework version upgrades are postponed indefinitely, new features bypass the system instead of integrating into it, and the team prefers to duplicate a component rather than modify the existing one.
In 2024 and 2025, the obligations under NIS2 and DORA hit systems that had not evolved in years head-on. The requirements for enhanced traceability, incident notification, and resilience testing presuppose a system capable of absorbing these changes without a complete rewrite. A frozen system cannot do this.
Security and compliance are part of ongoing maintenance
In many organizations, maintenance is still viewed through a corrective lens: application patches, incident resolution, availability monitoring. Security and compliance are treated as separate concerns, managed by other teams, on other cycles.
This separation creates a dangerous gap. Vulnerabilities in dependencies accumulate between two annual security reviews. Access rights drift with arrivals and departures. Technical logs do not cover what an auditor would expect. And when the audit arrives, the team spends more time reconstructing evidence than demonstrating control of the system.
Integrating security and compliance into ongoing maintenance means ensuring that dependency reviews, access rights checks, log quality, and sensitive change management are not one-off exercises but regular, tooled, and tracked activities.
Trust-condition maintenance requires a system-level view
Trust-condition maintenance requires looking at the application, infrastructure, identity, logs, data flows, dependencies, operations, and compliance as a whole. It is not a separate layer added on top of operational maintenance. It is a way of governing a system in production.
In practice, this translates into a few structuring elements:
- a living inventory of components and their dependencies;
- a vulnerability tracking dashboard with defined remediation timelines;
- periodic access rights reviews;
- regular restoration tests;
- a clear process for changes that affect sensitive components.
This also applies to AI systems
AI use cases do not escape this logic. They amplify it.
An AI system in production depends on models that evolve, connectors to data sources that change, access rights to sometimes sensitive corpora, and behaviors that are difficult to predict over time. A drifting model, a connector that changes format, a corpus that is no longer up to date: all situations where the system continues to respond, but quality degrades without any obvious warning signal.
Trust-condition maintenance for an AI system therefore requires, in addition to the standard foundation, continuous evaluation of output quality, monitoring of model and data drift, and clear rules for updates and rollback. Once again, the issue is not just about keeping the service available. It is about maintaining control.
Conclusion
Maintaining a system is not just about fixing what breaks. True maintenance means preserving the security, readability, compliance, and ability to evolve the system over time.
Keeping a production system available without losing control over its security, compliance and evolution. Managed services for trust-critical systems and modernization.