Skip to content
GCC AI Research

CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLMs

arXiv · · Significant research

LLM Research Ethics Product Policy

Summary

MBZUAI researchers introduced CopyShield, a new benchmark designed to compare various copyright defense mechanisms in large language models (LLMs) under controlled conditions. The benchmark evaluates three distinct intervention levels—contrastive decoding (output), Direct Preference Optimization (behavioral), and activation intervention (representation)—on LLaMA-3.1-8B and Mistral-7B-v0.3 models using public-domain books. Findings indicate that intervention levels are associated with distinct compliance-utility trade-offs, with DPO showing high degeneracy in LLaMA-3.1-8B while activation intervention effectively blocks non-literal queries before generation. Why it matters: This research provides crucial insights into developing more robust and ethically compliant LLMs by systematically evaluating methods to prevent unauthorized memorization and reproduction, a key concern for responsible AI deployment and adoption in the region.

Keywords

CopyShield · LLMs · Copyright · AI Ethics · Research Benchmark

Read original article →

Get the weekly digest

Top AI stories from the GCC region, every week.

Related

FAID: Fine-Grained AI-Generated Text Detection Using Multi-Task Auxiliary and Multi-Level Contrastive Learning

arXiv ·

MBZUAI researchers introduce FAID, a fine-grained AI-generated text detection framework capable of classifying text as human-written, LLM-generated, or collaboratively written. FAID utilizes multi-level contrastive learning and multi-task auxiliary classification to capture authorship and model-specific characteristics, and can identify the underlying LLM family. The framework outperforms existing baselines, especially in generalizing to unseen domains and new LLMs, and includes a multilingual, multi-domain dataset called FAIDSet.

M4: Multi-generator, Multi-domain, and Multi-lingual Black-Box Machine-Generated Text Detection

arXiv ·

MBZUAI researchers introduce M4, a multi-generator, multi-domain, and multi-lingual benchmark dataset for detecting machine-generated text. The study reveals challenges in generalizing detection across unseen domains or LLMs, with detectors often misclassifying machine-generated text as human-written. The dataset aims to foster research into more robust detection methods and is available on GitHub.