AI propaganda analysis reveals Llama 3 and Mistral model involvement

Researchers have developed a method to identify the AI models used in propaganda campaigns by analyzing text for specific linguistic patterns and prompt leaks. This technique was applied to the PROPAGIA corpus, which contains articles from the Storm-1516/CopyCop influence campaign. The analysis revealed that the propaganda articles were significantly more vague, subjective, and negative compared to human-written articles, and contained evidence of prompt instructions. The findings suggest that models from the Llama 3 family were likely used, with potential involvement of Mistral-family models. AI

IMPACT This research provides a method to detect AI-generated propaganda, potentially aiding in identifying and mitigating disinformation campaigns.

RANK_REASON The cluster contains an academic paper detailing a new forensic analysis method for AI-driven influence campaigns. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• INSIKT GROUP
• Llama 3
• Mistral AI
• PROPAGIA
• Sistema de Informação para o Património Arquitectónico
• Storm-1516/CopyCop
• VIGINUM

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →