Cloud environments face constant cybersecurity threats, requiring efficient and reliable defense mechanisms to ensure service continuity. This paper presents a novel approach leveraging AI agents to autonomously detect and mitigate threats in cloud systems, aligned with Site Reliability Engineering (SRE) practices. Using the NSL-KDD dataset, we train AI models to classify attack types accurately, achieving high precision and recall, particularly with Random Forest classifiers. We further simulate threat scenarios to measure the AI agent’s response times, comparing Time to Detect (TTD) and Time to Mitigate (TTM) against traditional methods. Results demonstrate a significant reduction in TTD and TTM, with the AI agent achieving up to 6x faster detection and mitigation. This autonomous capability not only improves threat response but also supports SRE-aligned metrics such as Mean Time to Detect (MTTD) and Mean Time to Resolve (MTTR), ensuring enhanced reliability in cloud infrastructures. By integrating AI-driven automation into cloud security operations, our findings underscore the potential of AI agents as proactive security operators, advancing both cybersecurity and operational resilience. This research contributes to the development of scalable, autonomous solutions crucial for the future of secure, resilient cloud computing.