MOSAIC introduces a novel approach for reliably detecting AI-generated content by leveraging multiple independent observers to improve detection robustness.

• Combines multiple detection signals from various sources to enhance reliability
• Addresses the growing security threat of LLM-generated harmful content
• Uses a classification framework that outperforms single-observer methods
• Provides a practical defense against increasingly sophisticated text generation

This research offers critical security capabilities as AI text becomes increasingly difficult to distinguish from human-written content, helping organizations identify potential forgeries and mitigate risks of synthetic content.

MOSAIC: Multiple Observers Spotting AI Content, a Robust Approach to Machine-Generated Text Detection