Today, 6G networks aspire to deliver unprecedented capabilities such as ultra-fast data speeds, sub-millisecond latency, extreme reliability, and ubiquitous global coverage, bringing wireless communications to the level of wired connections. These advances require fast and autonomous decision-making in the programming, control, and orchestration of end-to-end communication systems—requirements that traditional human-intervention network management techniques cannot meet. Consequently, there is a broad consensus that AI will permeate the entire 6G infrastructure.

Currently, AI models for mobile networks operate as closed, unreliable systems, which is a critical obstacle to their deployment in real networks. Agile-6G is a pioneer in the design of reliable Agentic AI systems adapted to 6G mobile networks. As defined by OpenAI, these systems are capable of pursuing complex goals with minimal human oversight. Unlike traditional AI models, which operate in isolation, Agentic AI systems integrate various learning techniques to achieve autonomy in decision-making, action planning, and learning, all subject to statements of intent established during development. In the 6G ecosystem, these intentions will define the goals, constraints, and safeguards that will guide AI-driven network control and optimization.

By introducing Agentic AI systems into 6G infrastructure, Agile-6G advances AI reliability and adaptability, ensuring that automation aligns with standards of reliability, transparency, and efficiency. This shift is imperative to harness the full potential of 6G, where AI will not just optimize individual components, but will function as an integrated, adaptable, and autonomous system capable of orchestrating network resources on an unprecedented scale.

Agile-6G project, PID2024-163089NB-I00, is funded by MICIU/AEI/10.13039/501100011033/ ERDF, EU.