The rise of AI-generated disinformation driven by large language models (LLMs) exacerbates threats to information trust and democracy. Motivated by the need for efficient detection, this paper integrates the DISARM framework with LLMs to enhance capabilities. The proposed OSINT LLM Agent leverages DISARM's structure and LLMs' reasoning for an effective, adaptable solution against disinformation campaigns. Experiments using real-world cases demonstrate feasibility and effectiveness, highlighting DISARM-LLM integration's potential for advanced detection.

PDF