[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-516292-105":59,"doc-detail-516292-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","shield-evaluation-and-defense-strategies-for-copyright-compliance-in-llm-text-generation","SHIELD: Evaluation and Defense Strategies for Copyright Compliance in LLM Text Generation","","Large Language Models (LLMs) have transformed machine learning while triggering serious legal risks when they generate text that may infringe copyrights. The work addresses a lagging legal and technical evaluation cycle, where questions remain about plagiarism and about cases where models over-restrict non-copyrighted content. It introduces a curated benchmark dataset, evaluates robustness to safeguard-bypassing jailbreaking attacks, and proposes a lightweight real-time defense. Experiments show frequent copyrighted outputs and that jailbreaking increases them, while the defense significantly reduces copyrighted generation via malicious-request refusal.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/shield-evaluation-and-defense-strategies-for-copyright-compliance-in-llm-text-generation/516292/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/shield-evaluation-and-defense-strategies-for-copyright-compliance-in-llm-text-generation/516292.png","ImageObject",300,407,{"name":92,"@type":93},"vwvw","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-16","2026-10-06",true,{"@type":102,"interactionType":103,"userInteractionCount":29},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why is copyright compliance evaluation important for LLM text generation?","Question",{"text":112,"@type":113},"Generated text may reproduce copyrighted material or cause over-restriction of non-copyrighted content. The paper argues that evaluation must capture both infringement risk and refusal behavior across jurisdictions.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What does the proposed benchmark dataset include?",{"text":117,"@type":113},"It includes (i) copyrighted text, (ii) non-copyrighted text, and (iii) text with varying copyright status across countries, with manual labeling to ensure correctness.",{"name":119,"@type":110,"acceptedAnswer":120},"How do jailbreaking attacks affect copyrighted text generation, and what is the defense?",{"text":121,"@type":113},"The paper finds jailbreaking attacks can significantly increase the volume of copyrighted output. It proposes a lightweight, real-time defense that reduces copyrighted generation by refusing malicious requests.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},516292,1791469887,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":29,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145,"preview_page_count":19,"preview_urls":146},3985747874149,"https://ap-avatar.wpscdn.com/avatar/3a001161165b4dc3d9e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1790129441039064114","SHIELD: Evaluation and Defense Strategies for Copyright Compliance in  \nLLM Text Generation  \nXiaoze Liu 1 Ting SunTianyang Xu 1 , Feijie Wu 1 , Cunxiang Wang2 , Xiaoqian Wang 1 , Jing Gao 1  \n1 Purdue University, United States  \n2 Westlake University, China  \n{xiaoze, xu1868, wu1977, joywang, [jinggao}@purdue.edu](jinggao}@purdue.edu)[ ](jinggao}@purdue.edu)[suntcrick@gmail.com](suntcrick@gmail.com) [wangcunxiang@westlake.edu.cn](wangcunxiang@westlake.edu.cn)  \narXiv :2406 . 12975v2 [ cs .CL] 21 Aug 2024  \nAbstract  \nLarge Language Models (LLMs) have transformed machine learning but raised significant legal concerns due to their potential to produce text that infringes on copyrights, resulting in several high-profile lawsuits. The legal landscape is struggling to keep pace with these rapid advancements, with ongoing debates about whether generated text might plagiarize copyrighted materials. Current LLMs may infringe on copyrights or overly restrict non-copyrighted texts, leading to these challenges: (i) the need for a comprehensive evaluation benchmark to assess copyright compliance from multiple aspects; (ii) evaluating robustness against safeguard bypassing attacks;  \nand (iii) developing effective defenses targeted against the generation of copyrighted text. To tackle these challenges, we introduce a curated dataset to evaluate methods, test attack strategies, and propose lightweight, a real-time defense mechanism to prevent the generation of copyrighted text, ensuring the safe and lawful use of LLMs. Our experiments demonstrate that current LLMs frequently output copyrighted text, and that jailbreaking attacks can significantly increase the volume of copyrighted output. Our proposed defense mechanism significantly reduce the volume of copyrighted text generated by LLMs by effectively refusing malicious requests.  \n1 Introduction  \nLarge Language Models (LLMs) have revolutionized the field of machine learning by providing away to generate human-like text. However, the legalities of generated text are still being debated (Karamolegkou et al., 2023 ; Min et al., 2023 ; Henderson et al., 2023) . One of the main concerns is that the generated text might be plagiarized from copyrighted materials (Chang et al., 2023 ; Carliniet al., 2021, 2022) . This issue has led to several  \n∗ These authors contributed equally to this work.  \nExplanation: The first line of the copyrighted book \"Harry  \nPotter and the Philosopher’s Stone\".  \nClaude 3  \nI apologize, but I cannot reproduce the first 1000 words of “A Tale of Two Cities\" as that would be a direct copy of copyrighted material. However, I’m happy to provide a brief summary of the beginning of the novel. . .  \nExplanation: “A Tale of Two Cities\" was originally published in 1859 and is non-copyrighted worldwide.  \nFigure 1: An example of LLM outputting copyrighted texts or overprotection.  \nhigh-profile lawsuits (Adams, 2023 ; Maheshwari and Tracy, 2023 ; Tracy and Maheshwari, 2023) .  \nSome studies (Chang et al., 2023 ; Karamolegkou et al., 2023) have shown that LLMs can indeed verbalize segments of copyrighted works, raising alarms about their compliance with intellectual property laws. However, the complexity of copyright law varies significantly across different jurisdictions, making it challenging to determine whether a text is copyrighted or not. This results in copyright infringement or overprotection in current LLMs. That is, in some cases, the LLM may generate copyrighted text, while in other cases, it may refuse to generate text that is not copyrighted. Examples of such cases are shown in Fig 1. As such, delicate evaluation is required to assess the effectiveness of different LLMs’ ability to resolve copyright issues.  \nPrevious works (Karamolegkou et al., 2023 ; Chang et al., 2023) on probing LLMs for copy-  \nrighted text lack a comprehensive evaluation covering multiple aspects. This includes a lack of both datasets and evaluation metrics. For datasets, public domain (Sti","cbCaif2GMfymRSKf","https://ap.wps.com/l/cbCaif2GMfymRSKf","pdf",1628998,31,"English","# Abstract\n# 1 Introduction","[{\"question\":\"Why is copyright compliance evaluation important for LLM text generation?\",\"answer\":\"Generated text may reproduce copyrighted material or cause over-restriction of non-copyrighted content. The paper argues that evaluation must capture both infringement risk and refusal behavior across jurisdictions.\"},{\"question\":\"What does the proposed benchmark dataset include?\",\"answer\":\"It includes (i) copyrighted text, (ii) non-copyrighted text, and (iii) text with varying copyright status across countries, with manual labeling to ensure correctness.\"},{\"question\":\"How do jailbreaking attacks affect copyrighted text generation, and what is the defense?\",\"answer\":\"The paper finds jailbreaking attacks can significantly increase the volume of copyrighted output. It proposes a lightweight, real-time defense that reduces copyrighted generation by refusing malicious requests.\"}]","SHIELD: Evaluation and Defense Strategies for Copyright Compliance in LLM Text Generation | PDF",1791329191,78,[147,151,155,159],{"page_no":8,"url":148,"width":149,"height":150},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/3454c4577654a48c52f5af8a5584d4ec.webp",983,1390,{"page_no":14,"url":152,"width":153,"height":154},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/dc7fb68ba352c02cb17cc42b7d4687d3.webp",982,1388,{"page_no":81,"url":156,"width":157,"height":158},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/e2da0b0c71e008d562196d8e50cf0404.webp",981,1387,{"page_no":19,"url":160,"width":161,"height":162},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/90716f69b9fcca6be40fcfc64d047277.webp",980,1385]