This research tackles a critical trust issue where API providers may secretly substitute premium LLMs with cheaper alternatives to reduce costs at users' expense.

• Identifies techniques to detect model substitution in black-box LLM APIs
• Demonstrates successful detection across multiple commercial providers
• Proposes verification mechanisms to ensure service integrity
• Reveals significant implications for benchmarking reliability and fair pricing

For security professionals, this work highlights urgent concerns about transparency in AI services and provides practical approaches to verify that customers receive the exact capabilities they're paying for.

Are You Getting What You Pay For? Auditing Model Substitution in LLM APIs