The growing social media use has caused a troubling issue of hate speech occurrence. As more individuals engage on these platforms, the harmful and offensive expressions towards certain groups have also escalated. Due to this, it is imperative to analyse this multi-modal data available online and eliminate toxic data to curb hate crimes on a global level. This book chapter presents an overview of multi-modal hate speech detection and publicly available datasets, followed by a discussion about the effectiveness of various machine and deep learning techniques used for multi-modal hate speech detection. Social media platforms now leverage AI technologies to filter out toxic content and combat hate crimes. Machine and deep learning techniques are gaining traction for analysing such data. The data collected on such social media sites include text, visual, and audio, leading to multi-modal data collection. Multi-modal data is used for improved accuracy and adaptability to get better results. Thus, the survey in this research delves into definitions of discriminatory speech, the rationale for discovery, and standard written content analysis strategies. It explores cutting-edge hate speech identification methods, including multi-modal ones, discussing their advantages and drawbacks. The paper also highlights datasets, their challenges, and their performance metrics and categorisation ratings of popular hate speech detection approaches. In conclusion, the paper provides insights into the current landscape, offering comparisons, addressing challenges, and proposing future research directions in multi-modal and multi-lingual hate speech detection, contributing to AI-driven social media analysis advancements.