CitePrism: Human-in-the-Loop AI for Citation Auditing and Editorial Integrity

This feed contains multiple arXiv papers with notable security implications: (1) "AI-Mediated Communication Can Steer Collective Opinion" demonstrates that LLMs systematically introduce directional biases when editing human messages and that such biases can be amplified across social networks — a vector for information operations and covert opinion manipulation. (2) "Reconstructing temporal multi-relational firm networks... semiconductor industry" shows an LLM+web-data OSINT methodology that can reconstruct detailed temporal supply‑chain, partnership, and ownership graphs for the semiconductor