Safety Evaluation

Definitions

1. The broader process of assessing whether an AI system behaves safely under expected use and misuse scenarios. Safety evaluation may include red teaming, benchmarks, policy review, and product-specific risk testing.

   In plain English: Evaluating whether an AI system behaves safely.

   Example: "The launch required a formal safety evaluation because the assistant could summarize sensitive customer documents."

Related Terms

• Red Teaming AI
• AI Red Team
• AI Metric
• AI Performance
• AI Testing
• Human Baseline
• LLM Benchmark
• Model Eval
• Model Safety
• Moderation
• Refusal
• Safety Benchmark
• Safety Policy
• Safety Testing
• Safety Training
• Truthfulness