Security in AI Manufacturing agents: Top 10 LLM attack prompts

This white paper outlines the ten most critical LLM security attacks against AI systems on the manufacturing plant floor, describing prompt injection, CoT forgery, multi-turn jailbreaks (e.g., Crescendo Attack), sycophancy, authorization flaws, and PII leakage patterns, and emphasizing safety, operational, security, and business/liability risks to industrial workflows.