Small language models for public administration: towards sustainable, trustworthy, and transparent AI systems

This article offers a brief overview of the opportunities and challenges associated with the adoption of artificial intelligence in Public Administration, taking as a reference the AILO case developed by the Municipality of Trieste, a specific example of the use of Large Language Models (LLMs) in a public consultation context. It highlights the main critical issues, including privacy protection, algorithmic opacity, dependence on external providers, and legal compliance. In this context, the paper explores the hypothesis of adopting Small Language Models (SLMs), trained on domain-specific datasets, which appear to offer advantages in terms of technical sustainability, institutional control, and alignment with the European regulatory framework (AI Act, GDPR). The aim is to contribute to the ongoing debate on the integration of AI into public processes by outlining criteria that may guide its development and adoption in a sustainable manner and in accordance with ethical and legal principles.