ICopyrightDetector<T>

Interfaces · AiDotNet.Safety.Text

Interface for copyright and memorization detection modules.

Implements: ITextSafetyModule<T>

Known implementations: CopyrightDetectorBase<T>

For Beginners

A copyright detector checks if an AI’s output copies from copyrighted books, articles, or code. It can detect when the AI is regurgitating memorized training data rather than generating original content.

How It Works

Copyright detectors identify potential copyright violations and training data memorization in model outputs. They use n-gram overlap analysis, embedding similarity to known works, and perplexity-based memorization detection.

References:

  • DE-COP: Detecting copyrighted content via paraphrased permutations (2024, arxiv:2402.09910)
  • Machine unlearning to remove memorized copyrighted content (2024, arxiv:2412.18621)
  • GPTZero: Hierarchical multi-task AI text detection (2026, arxiv:2602.13042)

Methods

Method Summary
double GetMemorizationScore(string text) Gets the memorization likelihood score (0.0 = original, 1.0 = memorized).