Agents4Gov: Privacy-Preserving Web Agents Using Open LLMs for Public Sector Tasks

• João Victor de Castro Oliveira USP
• Nathan Rufino USP
• Silvio Levcovitz USP / PGFN
• Matheus Yasuo Ribeiro Utino USP
• Fábio Lobato USP / UFOPA
• Marcelo Turine USP / UFMS
• Solange Rezende USP
• Antonio Fernando Lavareda Jacob Junior UEMA
• Ricardo Marcondes Marcacini USP

DOI: https://doi.org/10.5753/eniac.2025.13911

Resumo

Effective public governance requires an organizational culture that values transparency, technical competence, and ethical behavior, while also addressing persistent challenges related to efficiency and compliance in public sector institutions. The digital transformation of the public sector, particularly through the integration of Large Language Models (LLMs), represents a strategic shift toward data-driven governance, process automation, and scalable innovation. These emerging technologies offer unprecedented opportunities to enhance inclusiveness, transparency, security, efficiency, accessibility, and citizen-responsive public services. However, while closed-source LLMs can automate digital tasks, their adoption is limited by concerns over privacy and cost. We present Agents4Gov, a privacy-focused and cost-effective framework for automating browser tasks using open-source LLMs. Built on Browser-Use, it combines a powerful model for planning with a lightweight model for execution, supports asynchronous task requests via email, and can run on scheduled intervals to avoid the need for always-on infrastructure. Evaluated on MiniWoB++, Agents4Gov with LLaMA 4 models achieves competitive results compared to traditional Browser-Use agents powered by GPT-4o-based LLMs.

Referências

Alhosani, K. and Alhashmi, S. M. (2024). Opportunities, challenges, and benefits of ai innovation in government services: a review. Discover Artificial Intelligence, 4(1):18.

Aryatama, S., Miswan, M., Fahriyah, F., Pribadi, T., and Suacana, I. (2024). Enhancing governance efficiency through digital transformation in public services: Lessons from global practices. Global International Journal of Innovative Research, 2(5):1019–1027.

Brasil (2019). Decreto nº 9.901, de 8 de julho de 2019: Altera o decreto nº 9.203, de 22 de novembro de 2017, que dispõe sobre a política de governança da administração pública federal direta, autárquica e fundacional. Diário Oficial da União.

Browserbase (2024). Open operator: A template for building web agents with stagehand on browserbase. [link]. Accessed: Jun 20, 2025.

Cao, C., Zhuang, J., and He, Q. (2024). Llm-assisted modeling and simulations for public sector decision-making: Bridging climate data and policy insights. In AAAI-2024 Workshop on Public Sector LLMs: Algorithmic and Sociotechnical Design.

Convergence (2025). Convergence’s proxy ahead in top agent benchmark, beats openai and anthropic. [link]. Accessed: Jun 20, 2025.

Cordella, A. and Bonina, C. M. (2012). A public value perspective for ict enabled public sector reforms: A theoretical reflection. Government information quarterly, 29(4):512–520.

Hu, X., Xiong, T., Yi, B., Wei, Z., Xiao, R., Chen, Y., Ye, J., Tao, M., Zhou, X., Zhao, Z., et al. (2024). Os agents: A survey on mllm-based agents for computer, phone and browser use.

Li, X., Wang, S., Zeng, S., Wu, Y., and Yang, Y. (2024). A survey on llm-based multiagent systems: workflow, infrastructure, and challenges. Vicinagearth, 1(1):9.

Mamalis, M. E., Kalampokis, E., Fitsilis, F., and Tarabanis, K. (2024). A large language model agent based legal assistant for governance applications. In International Conference on Electronic Government, pages 286–301. Springer.

Melo, M. K. (2024). Implementing AI for enhanced public services gov.br: A methodology for the brazilian federal government. In Proceedings of the 20th International Conference on Web Information Systems and Technologies, WEBIST 2024, Porto, Portugal, November 17-19, 2024, pages 90–101.

Mergel, I., Dickinson, H., Stenvall, J., and Gasco, M. (2024). Implementing ai in the public sector. Public Management Review, pages 1–14.

Mudryi, M., Chaklosh, M., and Wójcik, G. (2025). The hidden dangers of browsing ai agents. arXiv preprint arXiv:2505.13076.

Musumeci, E., Brienza, M., Suriani, V., Nardi, D., and Bloisi, D. D. (2024). Llm based multi-agent generation of semi-structured documents from semantic templates in the public administration domain. In International Conference on Human-Computer Interaction, pages 98–117. Springer.

Müller, M. and Žunič, G. (2024). Browser use: Enable ai to control your browser.

Nardes, J. A. R., Altounian, C. S., and Vieira, L. A. G. (2018). Governança pública. O desafio do Brasil, 3.

OpenAI (2025). Introducing operator. [link]. Accessed: Jun 20, 2025.

S. B. LLC (2025). Do browser: Ai browser automation agent. [link]. Accessed: Jun 20, 2025.

Scutella, M., Plewa, C., and Reaiche, C. (2024). Virtual agents in the public service: examining citizens’ value-in-use. Public Management Review, 26(1):73–88.

Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., et al. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6):186345.

Wooldridge, M. (2009). An introduction to multiagent systems. John wiley & sons.

Wooldridge, M. and Jennings, N. R. (1995). Intelligent agents: Theory and practice. The knowledge engineering review, 10(2):115–152.