The escalating sophistication and volume of cyber threats present an unprecedented challenge to digital security, rendering traditional reactive defence mechanisms increasingly insufficient. This scholarly article explores the transformative role of Artificial Intelligence (AI) in revolutionising cybersecurity, offering proactive strategies to detect, mitigate, and even predict emerging threats. We begin by examining the evolving landscape of modern cyber threats, including ransomware, Advanced Persistent Threats (APTs), and particularly, the alarming rise of AI-powered offensive tactics. Subsequently, the paper delves into the diverse applications of AI in cyber defence, highlighting the capabilities of machine learning, deep learning, natural language processing, and reinforcement learning in enhancing threat detection, incident response, and vulnerability management. Critical analysis extends to the significant challenges inherent in AI-driven cybersecurity, such as adversarial AI, algorithmic bias, the explainability problem, and ethical implications. Finally, we discuss future directions, encompassing the convergence of AI with quantum computing, the imperative for human-AI collaboration, and the development of autonomous self-healing systems, alongside the necessary evolution of regulatory frameworks. This research ultimately posits that while AI is an indispensable tool for future cybersecurity, its effective and ethical deployment necessitates continuous innovation, rigorous oversight, and a balanced understanding of its immense potential and inherent risks.