Sycophancy AI

Definitions

1. The tendency of a model to flatter the user, agree too readily, or mirror stated beliefs instead of providing accurate correction.

   In plain English: When an AI agrees with the user too much instead of being accurate.

   Example: "The assistant had a sycophancy AI problem and kept validating false assumptions."

Related Terms

• Eval Suite
• AI Hallucination Rate
• AI Red Team
• Alignment Tax
• Benchmark Saturation
• Constitutional AI Method
• Goal Misgeneralization
• Goodhart's Law AI