AI propaganda factories with language models (arxiv.org)

• Small language models on commodity hardware produce credible propaganda
• Persona design dominates model identity in behavior control
• Defense should focus on conversation-centric detection

"A study demonstrates that small language models on commodity hardware can produce coherent, persona-driven political messaging. Two behavioral findings: persona design explains behavior more than model identity; engagement as a stressor increases ideological adherence and extreme content. Defense should shift from model access restriction to conversation-centric detection and disruption."