摘要
Abstract
The rise of generative artificial intelligence has reshaped content production,and in doing so,challenged traditional copyright audi-ting models built around the examination of completed works.Conventional auditing presumes identifiable authorship,clear rights provenance,and traceable creation processes.In the AIGC context,however,content is produced through complex and often opaque technical systems.Questions of subject qualification,the legality of training data,and the allocation of rights in generated outputs arise at a structural level,and cannot be adequately addressed through work-centered review alone.In response,the scope of copyright auditing must extend beyond individual outputs to the broader con-figuration of subject status,input rights chains,and output risk control.This shift entails moving from a model focused primarily on verifying results to one that examines the conditions under which generation occurs.Mechanisms such as risk allocation rules,red-flag testing,and the establishment of a"minimal evidence set"can provide a reasonably verifiable basis for assessing generative activities without requiring exhaustive technical scrutiny.By retaining information such as creative intention,generation records,and source data,a factual basis for copyright auditing can be provided.This provides a methodological reference for addressing the challenges of copyright auditing brought by AIGC.关键词
生成式人工智能/版权审计/过程不确定性/证据最小集/权利链Key words
Generative Artificial Intelligence/Copyright Auditing/Process Uncertainty/Minimum Evidence Set/Rights Chain分类
社会科学
