MiRAGeNews marks a significant advancement in efforts to combat AI-driven fake news. This comprehensive dataset comprises 12, 500 high-quality image-caption pairs, meticulously curated to include both authentic images and those generated by state-of-the-art AI technologies. Designed to aid researchers and developers, MiRAGeNews helps distinguish genuine news content from AI-created fake visuals that can mislead audiences and distort public perception. The rapid rise of AI-generated images and multimedia has introduced new difficulties in verifying news authenticity. Conventional fact-checking methods often struggle to keep up with sophisticated generative models that produce highly realistic visuals. Such AI-created fake visuals contribute to misinformation, propaganda, and diminished trust in legitimate news sources. To address this pressing issue, MiRAGeNews has been carefully assembled as a benchmark dataset for developing and evaluating detection systems that identify AI-generated content. By balancing 12, 500 image-caption pairs between real and synthetic images, the dataset ensures that detection algorithms can effectively learn to differentiate authentic visuals from counterfeit ones. Alongside the dataset, MiRAGeNews offers a multi-modal detector called MiRAGe. This tool harnesses the synergy between images and their captions to improve accuracy in detecting fake images.

Unlike traditional detectors focusing solely on either visual or textual data, MiRAGe’s approach exploits inconsistencies and subtle discrepancies between images and captions commonly found in AI-generated content. MiRAGe establishes a new standard by exceeding the detection performance of existing baseline models. This advancement is essential for media outlets, fact-checkers, and social media platforms committed to maintaining the integrity of publicly shared information. Incorporating MiRAGe into verification workflows enables more effective filtering of misleading AI-generated visuals before they can spread widely. Furthermore, making MiRAGeNews publicly available empowers the broader research community to innovate within this field. Researchers can use the dataset to develop, test, and refine detection models, fostering collaboration and accelerating progress in combating visual misinformation. The challenges posed by AI-generated fake news are complex, extending beyond technical issues to societal impacts such as undermining democratic discourse, increasing polarization, and eroding trust in journalism. Tools like MiRAGeNews and MiRAGe play a crucial role in a multifaceted response that also involves public awareness, policy development, and cross-sector cooperation. In summary, MiRAGeNews offers a vital resource suited to today’s media environment, where AI-generated content is increasingly sophisticated and prevalent. By combining a large, well-curated dataset with an advanced multi-modal detection system, it provides a promising means to mitigate the dangers of AI-driven misinformation. This initiative highlights the ongoing need for innovation and vigilance to preserve news credibility in the digital age.