Crisis Management Days

Implications of inadequate logging exploitation by malicious artificial intelligence agents

Tomislav Gligora

University of Applied Sciences of Velika Gorica

Igor Znika

Unifiedpost Solutions d.o.o.

Keywords: Application logging, Cybersecurity vulnerabilities, Malicious AI agents, Anomaly detection, Logging infrastructure

---

Abstract

Robust application logging is a critical component for securing systems. However poor logging practices, such as inconsistent event tracking and fragmented data, can introduce significant vulnerabilities within cybersecurity frameworks. Malicious Artificial Intelligence agents exploit gaps in system observability created by these deficiencies, hindering effective detection and response to threats. This paper explores the implications of inadequate logging infrastructure on system accountability and its potential for exploitation by Artificial Intelligence driven threats. By incorporating research from the domains of cybersecurity, artificial intelligence, anomaly detection, and digital forensics, it emphasizes the risks posed by inadequate logging in environments increasingly reliant on autonomous systems. Furthermore, this paper proposes strategies to enhance logging infrastructure to mitigate these risks, including real time anomaly detection, semantic log structuring, and improved auditability.

---

References

AI: Advent of Agents Opens New Possibilities for Attackers (2025.) Accessed: March 13, 2025: https://www.security.com/threat-intelligence/ai-agent-attacks

Cybercriminals Are Targeting AI Agents and Conversational Platforms: Emerging Risks for Businesses and Consumers (20024) Accessed: October 9, 2024: https://www.resecurity.com/blog/article/cybercriminals-are-targeting-ai-agents-and-conversational-platforms-emerging-risks-for-businesses-and-consumers