In this episode, we continue our deep dive into the evolution of artificial intelligence systems, focusing on a distinction that is becoming increasingly central in both regulatory and technological debates: the difference between General-Purpose AI (GPAI) models, Large Language Models (LLMs), and Agentic AI.
The growing overlap among these three levels—regulatory category, enabling technology, and autonomous architecture—makes it necessary to explore them in order to correctly interpret the obligations under the AI Act, assess the organizational impact of new models, and guide informed AI adoption in business and institutional processes.
The following analysis aims to distinguish what falls under the legal notion of GPAI, what belongs to the technological domain of LLMs, and what characterizes agentic systems, which possess increasing autonomy and can directly affect the real world.
We can imagine a hierarchy:
- GPAI represents the broadest regulatory and functional category,
- LLMs are a specific technology often included within that category, and
- Agentic AI is an architectural evolution that uses LLMs to act autonomously.
