Phishing attacks pose a significant cybersecurity risk, exploiting human trust to deceive users and steal confidential information. Traditional rule-based detection methods have proven ineffective against the constantly evolving strategies of cybercriminals. As a result, ML techniques have gained prominence in phishing detection. Prior studies indicate that Random Forest (RF) and SVM are one of the most significant reliable models due to their high accuracy and ability to minimize false positives. These models effectively analyze various phishing indicators, such as URLs, email content, and website characteristics. Despite these advancements, a major challenge remains most phishing detection systems are trained primarily on English-language datasets. Cybercriminals increasingly exploit this gap by designing phishing attacks in regional languages, evading detection by conventional models. This paper emphasizes the need for multilingual phishing detection by incorporating multilingual datasets. An ML-based framework is designed to enhance phishing detection across diverse linguistic contexts, strengthening cybersecurity in a globally interconnected digital landscape. A diverse dataset featuring text samples was utilized for training and evaluating the model. The results indicate that SVM is the most effective model, achieving an accuracy of 98.12%. The findings emphasize the significance of multilingual phishing detection in enhancing cybersecurity measures on a global scale.