Fantastic Copyrighted Beasts and How (Not) to Generate Them

2024-06-20

arXiv.org

AI Summary

This study systematically assesses the risk of copyright-infringing character reproduction—e.g., iconic video game characters—in image/video generative models, a critical yet underexplored AIGC copyright concern. Method: To address the absence of dedicated benchmarks, we introduce CopyCat, the first evaluation suite targeting copyrighted characters, featuring a dual-dimensional assessment framework (visual similarity + prompt alignment) and a semi-automatic semantic trigger word identification technique. Contribution/Results: Empirical analysis reveals that generic prompts (e.g., “videogame, plumber”) reliably elicit protected characters; conventional safety mechanisms like prompt rewriting alone are ineffective—only when combined with negative prompting do they suppress unauthorized generation. We further propose a deployable, multi-layered mitigation strategy that substantially reduces unauthorized character synthesis while preserving generation fidelity, offering both theoretical insights and practical guidelines for AIGC copyright governance.

Technology Category

• Humans and AI: Game Design — Virtual Humans, NPCs and Autonomous Characters
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Recent studies show that image and video generation models can be prompted to reproduce copyrighted content from their training data, raising serious legal concerns around copyright infringement. Copyrighted characters, in particular, pose a difficult challenge for image generation services, with at least one lawsuit already awarding damages based on the generation of these characters. Yet, little research has empirically examined this issue. We conduct a systematic evaluation to fill this gap. First, we build CopyCat, an evaluation suite consisting of diverse copyrighted characters and a novel evaluation pipeline. Our evaluation considers both the detection of similarity to copyrighted characters and generated image's consistency with user input. Our evaluation systematically shows that both image and video generation models can still generate characters even if characters' names are not explicitly mentioned in the prompt, sometimes with only two generic keywords (e.g., prompting with"videogame, plumber"consistently generates Nintendo's Mario character). We then introduce techniques to semi-automatically identify such keywords or descriptions that trigger character generation. Using our evaluation suite, we study runtime mitigation strategies, including both existing methods and new strategies we propose. Our findings reveal that commonly employed strategies, such as prompt rewriting in the DALL-E system, are not sufficient as standalone guardrails. These strategies must be coupled with other approaches, like negative prompting, to effectively reduce the unintended generation of copyrighted characters. Our work provides empirical grounding to the discussion of copyright mitigation strategies and offers actionable insights for model deployers actively implementing them.

Problem

• Assessing ease of generating copyrighted characters unintentionally
• Evaluating effectiveness of existing copyright mitigation strategies
• Developing framework to measure similarity and intent consistency

Innovation

• Novel evaluation framework for copyrighted content
• Semi-automatic keyword identification techniques
• Enhanced mitigation strategies including negative prompting