Generative Artificial Intelligence is a branch of AI that focuses on creating new content based on initial inputs, known as prompts.
It relies on models capable of identifying fundamental structural patterns in large volumes of data (texts, images, sounds, etc.) to generate plausible and coherent content.

This approach has seen rapid growth in recent years, driven by advances in deep learning and the increasing availability of powerful models, such as large language models (LLMs)—like OpenAI’s ChatGPT—or image generators like DALL·E or Stable Diffusion.

Some concrete applications of generative AI include:

• Text generation (assisted writing, code generation, automated replies, etc.);

• Image creation from textual descriptions (useful in advertising, fashion, artistic creation, etc.);

• Data augmentation or synthetic data generation to train other models;

• Autonomous agents capable of gathering information and interacting with their environment to achieve specific goals (such as automatically handling customer service, generating and testing software code, or automating competitive monitoring).

Proprietary language model providers generally offer convenient APIs with pay-per-use pricing. However, these solutions raise significant concerns regarding data privacy and digital sovereignty.

As an alternative, open-weight models, deployable locally or through sovereign operators, offer:

• complete control over data;
• better regulatory compliance;
• adaptability to business needs;
• and reduced technological dependency.

CATIE supports public and private stakeholders in selecting and adopting the most relevant technologies according to their needs, ensuring a balance between performance, sovereignty, control over usage, and budget.