GUARDING THE GUARDIAN: AI-ENABLED THREAT DETECTION AND PREVENTION WITH SAFEGUARDS FOR AGENTIC AND ADVERSARIAL RISKS

There is a possibility to improve the process of detecting and preventing cybersecurity threats by means of the analysis of large security data sets, pattern recognition, anomalies detection, unstructured threat processing, prioritization, and incident response with the help of artificial intelligence technology. There are different capabilities of different approaches, like supervised learning, anomaly detection, deep learning, NLP, and LLM. Also, the article demonstrates that AI cannot be used as an independent security solution per se. Model errors, limitations of data, adversarial attacks, prompt injection, data poisoning, information leakage, and agency problems could lead to failures. It is necessary to protect both systems that are protected from threats and AI elements that perform the protection. The suggested framework is based on telemetry, secure data engineering, multi-model analysis, contextual risk scoring, human/policy validation, controlled prevention, and continuous assurance. There is also a necessity to use the additional controls for AI agents, like sandboxing, instruction isolation, limited privileges, rate limiting, and approval gates. On the whole, the most sensible way forward is to have AI as a tool for enhancing the cybersecurity operations in a controlled way. By integrating AI into the concept of defense-in-depth, ensuring a secure lifecycle of models and proper governance along with the right kind of human oversight, AI can be used effectively in the realm of security.

1.   Kaur, R., Gabrijelčič, D., Klobučar, T.: Artificial intelligence for cybersecurity: Literature review and future research directions. Information   Fusion,    97,  101804 (2023). https://doi.org/10.1016/j.inffus.2023.101804 2.   Sommer, R., Paxson, V.: Outside the Closed World: On Using Machine Learning for Network Intrusion Detection. IEEE Symposium on Security and Privacy, 305–316 (2010). https://doi.org/10.1109/SP.2010.25 3.   Vassilev, A., Oprea, A., Fordyce, A., Anderson, H., Davies, X., Hamin, M.: Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations. NIST AI 100-2 E2025 (2025). https://doi.org/10.6028/NIST.AI.100-2e2025 4.   OWASP Gen AI Security Project: OWASP Top 10 for Large Language Model Applications 2025. Open Worldwide Application  Security    Project (2025). https://genai.owasp.org/llm-top-10/ 5.  Buczak, A.L., Guven, E.: A Survey of Data Mining and Machine Learning Methods for Cyber Security Intrusion Detection. IEEE Communications Surveys & Tutorials, 18(2), 1153–1176 (2016). https://doi.org/10.1109/COMST.2015.2494502 6.   Shone, N., Nguyen, T.N., Phai, V.D., Shi, Q.: A Deep Learning Approach to Network Intrusion Detection. IEEE Transactions on Emerging Topics in Computational Intelligence, 2(1), 41–50 (2018). https://doi.org/10.1109/TETCI.2017.2772792 7.   Vinayakumar, R., Alazab, M., Soman, K.P., Poornachandran, P., Al-Nemrat, A., Venkatraman, S.: Deep Learning Approach for Intelligent Intrusion Detection System. IEEE Access, 7, 41525–41550(2019). https://doi.org/10.1109/ACCESS.2019.2895334 8.  Ferrag, M.A., Maglaras, L., Moschoyiannis, S., Janicke, H.: Deep Learning for Cyber Security Intrusion Detection: Approaches, Datasets, and Comparative Study. Journal of Information Security and   Applications,                 50,     102419     (2020). https://doi.org/10.1016/j.jisa.2019.102419 9.  Apruzzese, G., Laskov, P., Montes de Oca, E., et al.: The Role of Machine Learning in Cybersecurity. Digital Threats: Research and Practice, 4(1), Article 8 (2023). https://doi.org/10.1145/3545574 10. Chen, Y., Cui, M., Wang, D., Cao, Y., Yang, P., Jiang, B., Lu, Z., Liu, B.: A Survey of Large Language Models for Cyber Threat Detection. Computers & Security, 145, 104016 (2024). https://doi.org/10.1016/j.cose.2024.104016