Google has announced the release of DataGemma, a series of open models designed to enhance the factual accuracy of large language models (LLMs). DataGemma utilizes real-world data from Data Commons through two approaches: Retrieval Interleaved Generation (RIG) and Retrieval Augmented Generation (RAG). These methods help to ground LLMs in reliable, public statistical data, thereby reducing instances of hallucinations in AI responses. The models, including the fine-tuned Gemma 2 and a 27B-sized version, are aimed at improving responsible AI development and fact-checking capabilities.