This research addresses the critical challenge of identifying the original models behind black-box LLMs that may be using proprietary technology without authorization.

• Introduces PlugAE, a novel technique designed to trace and identify the origin of black-box LLMs
• Highlights how third parties can conceal unauthorized LLM usage through fine-tuning and limited API access
• Demonstrates practical methods to detect unfair competition and licensing violations
• Provides a technical solution for model attribution and external auditing

For security professionals, this research offers essential tools to protect intellectual property in the rapidly growing LLM market, enabling proper attribution and enforcement of licensing agreements.

The Challenge of Identifying the Origin of Black-Box Large Language Models