This survey provides an overview of generative AI methods in de novo drug design, focusing on the creation of novel biological compounds from scratch. The survey is organized into two main themes: small molecule and protein generation, with additional sections on antibody and peptide generation. It highlights the rapid development and complexity of the field, making it challenging for new researchers to enter. The survey covers various subtasks and applications, including datasets, benchmarks, and model architectures, and compares the performance of top models. It discusses the challenges and approaches in both fields, emphasizing the potential of AI-driven de novo drug design to transform synthetic biology and drug discovery. The survey also provides a broad perspective on the field, highlighting relationships between different areas and future directions. Key models and their performance are detailed, with a focus on diffusion models, variational autoencoders, and generative adversarial networks (GANs). The survey concludes with a comprehensive summary of current trends, top-performing models, and future directions in ML-driven de novo drug design.This survey provides an overview of generative AI methods in de novo drug design, focusing on the creation of novel biological compounds from scratch. The survey is organized into two main themes: small molecule and protein generation, with additional sections on antibody and peptide generation. It highlights the rapid development and complexity of the field, making it challenging for new researchers to enter. The survey covers various subtasks and applications, including datasets, benchmarks, and model architectures, and compares the performance of top models. It discusses the challenges and approaches in both fields, emphasizing the potential of AI-driven de novo drug design to transform synthetic biology and drug discovery. The survey also provides a broad perspective on the field, highlighting relationships between different areas and future directions. Key models and their performance are detailed, with a focus on diffusion models, variational autoencoders, and generative adversarial networks (GANs). The survey concludes with a comprehensive summary of current trends, top-performing models, and future directions in ML-driven de novo drug design.