The field of natural language processing is witnessing a significant shift with the rapid advancement of large language models (LLMs). Researchers are now focusing on developing methods to detect and analyze AI-generated text, which has become increasingly sophisticated and challenging to distinguish from human-written content. One of the key areas of research is the development of features and classifiers that can effectively identify LLM-generated text. Studies have shown that certain features, such as those derived from the NELA toolkit, can capture nuanced linguistic and stylistic differences between human-written and AI-generated text. Another important direction is the investigation of the impact of LLMs on the spread of disinformation and the potential for machine-generated content to be used in malicious ways. Empirical evidence has confirmed the presence of LLM-generated text in real-world disinformation datasets, highlighting the need for more effective detection and mitigation strategies. The use of synthetic data generated by LLMs is also being explored as a means to improve language detection tasks, such as inclusive language detection and fake news classification. Initial results have shown promise, with fine-tuned models trained on synthetic data outperforming those trained on real data in some cases. Noteworthy papers include:

• A study on the use of natural language features for AI-generated text detection, which found that NELA features outperform RAIDAR features in both binary and multi-class classification tasks.
• Research on the growing presence of LLM-generated texts in multilingual disinformation, which provided empirical evidence of the increase in machine-generated content following the release of ChatGPT.
• A novel methodology for generating synthetic fake news through fact-based manipulations using LLMs, which demonstrated the effectiveness of transformer models in leveraging synthetic data for fake news detection.