<![CDATA[The development of Generative Artificial Intelligence (Generative AI), particularly Large Language Models (LLM), has opened new opportunities and simultaneously created risks in computer security. This study aims to analyze the utilization, opportunities, risks, and mitigation strategies of Generative AI in the context of computer security. The study used a qualitative approach with a literature review method of journal articles, proceedings, books, and relevant scientific reports from the period 2021–2026. The literature was searched through Google Scholar, IEEE Xplore, ScienceDirect, SpringerLink, and the ACM Digital Library, then analyzed using content analysis. The results of the study indicate that Generative AI has the potential to support threat detection, cyberthreat intelligence, malware analysis, penetration testing, code security, phishing analysis, and incident response. However, this technology also has dual-use characteristics because it can be used for phishing, social engineering, malicious code modification, and presents risks of prompt injection, data poisoning, information leakage, hallucinations, and over-reliance on AI. Mitigation requires human validation, the principle of least privilege, access control, data protection, and structured AI governance. Thus, Generative AI should be positioned as a decision-support technology that enhances the capabilities of security professionals, rather than replacing human oversight entirely.]]>
