AI-Generated Image Detection: Challenges and Recent Advances

Recent advances in the field of AI-based content generation in combination with the easy and widespread access to such tools can enhance disinformation campaigns with respect to plausibility and scale. This poses a significant threat to online information integrity, social trust, and even democratic processes. The rapid evolution of generative models demands robust and adaptable detection mechanisms. In this chapter, we outline the progress in the field of AI-Generated Image Detection (AIGID) including recent methods, relevant datasets, and assessment of existing methods’ performance. In addition, we discuss the challenges that the corresponding methods face such as performance drop in the wild due to post-processing operations, generalization to data from new generators, and transparency of detection workflow.