Open Access
ARTICLE
Agentic Artificial Intelligence in Cybersecurity: Emerging Risks, Governance Challenges, and Defensive Frameworks
Issue Vol. 2 No. 02 (2026): April May --- Section Articles --- Published Date: 2026-05-27
Abstract
Agentic artificial intelligence is emerging as a new class of autonomous AI systems capable of reasoning, planning, tool use, information retrieval, and independent task execution across digital environments. In cybersecurity, these capabilities create major opportunities for threat detection, vulnerability assessment, incident triage, and automated response. However, they also introduce new risks because agentic AI systems may access sensitive data, interact with APIs, execute security-related actions, and operate with limited human supervision. This article examines the cybersecurity risks, governance challenges, and defensive frameworks associated with agentic AI deployment. It highlights key concerns such as prompt injection, unauthorized tool use, data leakage, insecure automation, supply-chain exposure, weak accountability, and insufficient runtime monitoring. Drawing on AI risk management guidance, generative AI governance principles, zero trust architecture, and research on tool-connected language models, the article argues that secure agentic AI adoption requires a layered defensive approach. This includes agent identity control, least privilege access, secure tool governance, prompt and input validation, behavioral monitoring, audit logging, explainability, and human oversight for high-risk actions. The article concludes that agentic AI can strengthen cybersecurity operations, but only when deployed within a transparent, accountable, and continuously monitored governance framework that balances autonomy with organizational control (NIST, 2023, 2024; OWASP Gen AI Security Project, 2025; Schick et al., 2023; Yao et al., 2022).
