We propose a new active learning approach using Generative Adversarial Networks (GAN). Different from regular active learning, we adaptively synthesize training instances for querying to increase learning speed. Our approach outperforms random generation using GAN alone in active learning experiments. We demonstrate the effectiveness of the proposed algorithm in various datasets when compared to other algorithms. To the best our knowledge, this is the first active learning work using GAN. Generative Adversarial Active Learning