LLM Security

Definitions

1. The security discipline around large language model systems, including prompt injection defenses, data protection, access control, abuse prevention, and safe tool integration. LLM security extends beyond ordinary API security because model behavior itself can be influenced adversarially.

   In plain English: The security practices needed to protect large language model systems.

   Example: "The review focused on LLM security because the assistant could read sensitive tickets and trigger downstream actions."

Related Terms

• System Prompt Extraction
• Jailbreak AI
• LLM Benchmark
• LLM Chain
• LLM Gateway
• LLM Judge
• LLM Observability
• Prompt Attack
• Red Team AI