Assessing the Impact of AI TRiSM (Trust, Risk, and Security Management) ModelOps on the Operational Behavior of Generative AI Systems
Resumo
Este trabalho avalia o impacto de mecanismos ModelOps inspirados no AI TRiSM em um sistema de IA generativa local baseado em um pequeno modelo de linguagem (Small Language Model, SLM) e Ollama. Dois ambientes foram comparados: um sem controles de governança e outro incorporando mecanismos de monitoramento, auditoria e aplicação de políticas. Um teste de benchmark com 300 prompts foi executado em ambos os ambientes. Os resultados revelaram diferenças estatisticamente significativas entre as métricas operacionais avaliadas, demonstrando que os mecanismos ModelOps melhoram a eficácia da governança, a rastreabilidade, as capacidades de monitoramento e o controle operacional em sistemas de IA generativa, embora ao custo de um aumento na sobrecarga computacional.Referências
Bender, E. M., Gebru, T., McMillan-Major, A., and Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT ’21, pages 610–623, New York, NY, USA. Association for Computing Machinery.
Bommasani, R., Kapoor, S., Klyman, K., Longpre, S., Ramaswami, A., Zhang, D., Schaake, M., Ho, D. E., Narayanan, A., and Liang, P. (2024). Considerations for governing open foundation models. Science, 386(6718):151–153.
Chicco, D., Sichenze, A., and Jurman, G. (2025). A simple guide to the use of Student’s t-test, Mann-Whitney U test, Chi-squared test, and Kruskal-Wallis test in biostatistics. BioData Mining, 18(1):56.
Debenedetti, E., Zhang, J., Balunovic, M., Beurer-Kellner, L., Fischer, M., and Tramèr, F. (2024). AgentDojo: A dynamic environment to evaluate prompt injection attacks and defenses for LLM agents. In Advances in Neural Information Processing Systems, volume 37, pages 82895–82920.
Greshake, K., Abdelnabi, S., Mishra, S., Endres, C., Holz, T., and Fritz, M. (2023). Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection. In Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security, pages 79–90. ACM.
Grönman, J., Pohjola, T., and Haukioja, T. (2025). Cybersecurity challenges in open data and AI systems for sustainable local development. In 2025 MIPRO 48th ICT and Electronics Convention, pages 1088–1093.
Litan, A. (2026). AI trust and AI risk: Tackling trust, risk and security in AI models. Available at: [link]. Accessed: 2026-06-20.
Matotek, K., Cassel, H., Amiruzzaman, M., and Ngo, L. B. (2025). Evaluating the limitations of local LLMs in solving complex programming challenges.
Ray, P. P. (2026). A review of TRiSM frameworks in artificial intelligence systems: Fundamentals, taxonomy, use cases, key challenges and future directions. Expert Systems, 43(3):e70213.
Raza, S., Sapkota, R., Karkee, M., and Emmanouilidis, C. (2026). TRiSM for agentic AI: A review of trust, risk, and security management in LLM-based agentic multi-agent systems. AI Open, 7:71–95.
Romano, J., Kromrey, J. D., Coraggio, J., and Skowronek, J. (2006). Appropriate statistics for ordinal level data: Should we really be using t-test and cohen’s d for evaluating group differences on the nsse and other surveys? Annual Meeting of the Florida Association of Institutional Research, pages 1–33.
Shahin, N. and Alsmadi, I. (2026). Benchmarking Llama model security against OWASP top 10 for LLM applications. arXiv:2601.19970v1.
Trusted-AI (2024). Adversarial robustness toolbox (art). [link]. Accessed: 2026-08-13.
Wang, W.-C. (2024). Legal, policy, and compliance issues in using AI for security: Using Taiwan’s cybersecurity management act and penetration testing as examples. In 2024 16th International Conference on Cyber Conflict: Over the Horizon (CyCon), pages 161–176.
Zhang, C., Jin, M., Yu, Q., Liu, C., Xue, H., and Jin, X. (2024). Goal-guided generative prompt injection attack on large language models. In 2024 IEEE International Conference on Data Mining (ICDM), pages 941–946. IEEE.
Zou, A., Wang, Z., Carlini, N., Nasr, M., Kolter, J. Z., and Fredrikson, M. (2023). Universal and transferable adversarial attacks on aligned language models.
Bommasani, R., Kapoor, S., Klyman, K., Longpre, S., Ramaswami, A., Zhang, D., Schaake, M., Ho, D. E., Narayanan, A., and Liang, P. (2024). Considerations for governing open foundation models. Science, 386(6718):151–153.
Chicco, D., Sichenze, A., and Jurman, G. (2025). A simple guide to the use of Student’s t-test, Mann-Whitney U test, Chi-squared test, and Kruskal-Wallis test in biostatistics. BioData Mining, 18(1):56.
Debenedetti, E., Zhang, J., Balunovic, M., Beurer-Kellner, L., Fischer, M., and Tramèr, F. (2024). AgentDojo: A dynamic environment to evaluate prompt injection attacks and defenses for LLM agents. In Advances in Neural Information Processing Systems, volume 37, pages 82895–82920.
Greshake, K., Abdelnabi, S., Mishra, S., Endres, C., Holz, T., and Fritz, M. (2023). Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection. In Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security, pages 79–90. ACM.
Grönman, J., Pohjola, T., and Haukioja, T. (2025). Cybersecurity challenges in open data and AI systems for sustainable local development. In 2025 MIPRO 48th ICT and Electronics Convention, pages 1088–1093.
Litan, A. (2026). AI trust and AI risk: Tackling trust, risk and security in AI models. Available at: [link]. Accessed: 2026-06-20.
Matotek, K., Cassel, H., Amiruzzaman, M., and Ngo, L. B. (2025). Evaluating the limitations of local LLMs in solving complex programming challenges.
Ray, P. P. (2026). A review of TRiSM frameworks in artificial intelligence systems: Fundamentals, taxonomy, use cases, key challenges and future directions. Expert Systems, 43(3):e70213.
Raza, S., Sapkota, R., Karkee, M., and Emmanouilidis, C. (2026). TRiSM for agentic AI: A review of trust, risk, and security management in LLM-based agentic multi-agent systems. AI Open, 7:71–95.
Romano, J., Kromrey, J. D., Coraggio, J., and Skowronek, J. (2006). Appropriate statistics for ordinal level data: Should we really be using t-test and cohen’s d for evaluating group differences on the nsse and other surveys? Annual Meeting of the Florida Association of Institutional Research, pages 1–33.
Shahin, N. and Alsmadi, I. (2026). Benchmarking Llama model security against OWASP top 10 for LLM applications. arXiv:2601.19970v1.
Trusted-AI (2024). Adversarial robustness toolbox (art). [link]. Accessed: 2026-08-13.
Wang, W.-C. (2024). Legal, policy, and compliance issues in using AI for security: Using Taiwan’s cybersecurity management act and penetration testing as examples. In 2024 16th International Conference on Cyber Conflict: Over the Horizon (CyCon), pages 161–176.
Zhang, C., Jin, M., Yu, Q., Liu, C., Xue, H., and Jin, X. (2024). Goal-guided generative prompt injection attack on large language models. In 2024 IEEE International Conference on Data Mining (ICDM), pages 941–946. IEEE.
Zou, A., Wang, Z., Carlini, N., Nasr, M., Kolter, J. Z., and Fredrikson, M. (2023). Universal and transferable adversarial attacks on aligned language models.
Publicado
19/10/2026
Como Citar
FERNANDES, Jônatas; REBOUÇAS, Emanuel; XIMENES, Erisvan; PINTO, Hywre César de Brito; RODRIGUES, Emanuel B..
Assessing the Impact of AI TRiSM (Trust, Risk, and Security Management) ModelOps on the Operational Behavior of Generative AI Systems. In: ENCONTRO NACIONAL DE INTELIGÊNCIA ARTIFICIAL E COMPUTACIONAL (ENIAC), 23. , 2026, Cuiabá/MT.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 1828-1839.
ISSN 2763-9061.
DOI: https://doi.org/10.5753/eniac.2026.31411.
