AI Safety vs AI Security in LLM Applications: What Teams Must Know
Summary
This article distinguishes AI safety (protecting people from harmful outputs) from AI security (protecting systems from adversaries) and provides technical testing strategies using Promptfoo.
Key quotes
Safety protects people from your model's behavior. Security protects your LLM stack and data from adversaries.
If the model says something harmful, that's safety. If an attacker makes the model do something harmful, that's security.
The post details the technical differences between safety and security in LLMs, referencing 2024-2025 incidents including Replit and xAI. It includes practical configuration examples for automated testing using LLM-as-a-Judge rubrics.