Publications on information system modernization, digital trust and responsible AI industrialization.
No articles match your search.
Latest·6 min read
Data breaches: why MFA is now the sine qua non of trust.
Breaches keep coming, credentials keep circulating. In 2025, the right question is no longer how to protect your password: it is how to design authentication so that a stolen password has no operational consequence.
Open source and self-hosted AI models: real capabilities, real constraints
Llama, Mistral, Qwen, Gemma: open source models have considerably reduced the gap with frontier models. But 'free' does not mean no cost, and 'open source' does not mean no constraints. What you really need to know before choosing.
eIDAS 2 and the European e-wallet: what it concretely changes for your user journeys
The eIDAS 2 regulation and the European Digital Identity Wallet (EUDIW) will reshape identification and signing journeys. What CIOs need to anticipate right now.
New AI models in late 2025: what really changes for your production systems
Claude Opus 4.5, GPT-5, Gemini Ultra 2... Late 2025 was marked by an acceleration in frontier model releases. Beyond the benchmarks, what really changes for teams operating AI in real conditions.
Electronic signature: a practical guide to choosing the right eIDAS level
SES, AES, QES: three levels of electronic signature with very different evidential values and implementation constraints. A guide for choosing without over-engineering or under-dimensioning.
Trust Condition Maintenance: why it changes everything about application maintenance
Classic application maintenance answers the question 'does it work?'. Trust Condition Maintenance answers a different question: 'can it be exploited, betrayed or silently fail?'. These are not the same thing.
AI Act: what technical teams need to anticipate right now
The European AI regulation has entered into application. For teams developing or deploying AI systems, the concrete obligations around documentation, traceability and human oversight are closer than you think.
LLMOps: what it really means to put an AI model in production
Deploying an LLM is not like deploying a classic API. Governance, supervision, drift, rollback, cost: the reality of operating a language model in a critical information system.
Modular monolith vs microservices: why we so often choose the wrong one
Migration to microservices has become a reflex. It is often a mistake. The criteria that make a modular monolith the best answer, and the signals that indicate otherwise.
FIDO2 and Passkeys: What Production Deployment Actually Looks Like
Passkeys eliminate passwords, in theory. In practice, deploying FIDO2 in a production environment raises real questions about fallback, account recovery, and device management.
Before modernizing, you need to understand what you're modernizing. A practical guide to quantifying technical debt and defining a data-driven remediation roadmap.