CopyrightCatcher is an AI-powered tool developed by Patronus AI to detect instances where large language models (LLMs) reproduce copyrighted content verbatim. It is designed to help organizations identify and mitigate legal risks associated with unintended copyright infringement in AI-generated outputs.
Benefits & outcomes
In an adversarial test conducted by Patronus AI, several LLMs were evaluated for their propensity to reproduce copyrighted material. The results were as follows:
- OpenAI’s GPT-4: 44% of prompts resulted in copyrighted content reproduction.
- Mistral’s Mixtral-8x7B-Instruct-v0.1: 22%
- Meta’s Llama-2-70b-chat: 10%
- Anthropic’s Claude-2.1: 8%
These findings underscore the importance of tools like CopyrightCatcher in ensuring compliance and reducing potential legal liabilities
Notes & quotes
“Managing risks from unintended copyright infringement in LLM outputs should be a central focus for companies deploying LLMs in production.”
“LLM training data often contains copyrighted works, and it is pretty easy to get an LLM to generate exact reproductions from these texts.”
“It is critical to catch these reproductions, since they pose significant legal and reputational risks for companies that build and use LLMs in production systems.”
Publisher / company: Patronus AI
Date of case study: 06/03/2025
Source: https://www.patronus.ai/blog/introducing-copyright-catcher