Phishing attacks have become increasingly sophisticated, targeting users across multiple languages and platforms. These attacks increasingly exploit linguistic and cultural nuances, yet most detection systems focus on English, leaving low-resource languages like Russian and Uzbek vulnerable. To address this issue, we present UzPhishNet, a novel Graph Convolutional Network (GCN)-based model designed for multilingual phishing detection, with a focus on Russian and Uzbek languages. UzPhishNet achieves an accuracy of 99% on the Russian and Uzbek Phishing Dataset. The model leverages Dynamic Feature Importance Adjustment and K-Nearest Neighbors (KNN)-based graph construction to tackle key challenges such as data scarcity, language-specific characteristics, and the dynamic nature of phishing attacks. The model adjusts the importance of textual features during training using a Dynamic Feature Importance Adjustment mechanism. This mechanism prioritizes the most discriminative features, allowing the model to focus on features that contribute most to phishing detection, enabling it to focus on the most discriminative patterns for phishing detection. Additionally, the KNN-based graph construction captures relationships between text samples, enhancing the model’s ability to generalize across diverse phishing scenarios. To support this research, we introduce the Russian and Uzbek Phishing Dataset, a publicly available dataset comprising 100,000 samples evenly distributed between the two languages. The dataset is collected from diverse sources, including phishing emails, SMS messages, social media posts, and online forums, ensuring comprehensive coverage of phishing tactics in these languages. This dataset addresses the critical need for high-quality, language-specific resources in cybersecurity research. This work contributes to the growing body of research on cybersecurity by providing a scalable and effective solution for detecting phishing content in multilingual settings. By addressing the unique challenges of low-resource languages, UzPhishNet paves the way for future advancements in phishing detection and prevention.