Glossary
The key terms of enterprise AI, sovereignty and compliance, explained clearly, without jargon.
An AI system able to chain several actions autonomously to accomplish a goal: search, analyze, write, use tools.
The world's first regulatory framework for artificial intelligence, adopted by the European Union in 2024, with progressive application by risk level.
A computing environment physically or logically isolated from the internet, used for the most sensitive data and systems.
A 2018 US law allowing US authorities to demand access to data held by American companies, wherever it is hosted in the world.
Partial retraining of an AI model on specific data to specialize it for a domain, tone or task.
A design principle where a human validates the critical decisions or actions of an AI system before they are executed.
Artificial intelligence whose infrastructure, data and models are controlled by the company, without legal dependence on a foreign operator.
An AI model trained on huge text corpora, able to understand and generate natural language, the building block of AI assistants.
An AI model whose weights are published and downloadable, allowing it to run on your own infrastructure without depending on its creator's cloud.
A deployment mode where the software and data run on the company's servers, on its premises or in its datacenter, rather than in a provider's cloud.
A technique that connects an AI model to the company's documents: the AI retrieves the relevant passages before answering, with cited sources.
The application of the General Data Protection Regulation to processing carried out by artificial intelligence systems.
An ANSSI qualification standard guaranteeing that a cloud service offers the highest level of security and protection against extraterritorial laws.
Hosting an application yourself, including an AI platform, on infrastructure you control, rather than consuming a SaaS.
Use of AI tools by employees without the company's approval or oversight, exposing sensitive data to third-party services.