【摘要】 以Sora、Midjourney、Stable Diffusion为代表的文生图像类人工智能的多元应用，赋予人工智能生成图像训练数据侵权、版权冲突、知识产权定位等问题新的讨论意蕴。文生图像类人工智能涉及的知识产权侵犯风险包括训练数据构成潜在侵权、知识产权归属尚未明晰、商业应用扩大侵权规模三个方面。通过强化图像的独特性与可溯源性、完善侵权责任认定与承担规制、增进行业自律与版权知识普及等治理手段，可有效治理前述知识产权侵犯风险，共同推动文生图像类人工智能的合规、高效与可持续发展。更多还原

【Abstract】 The diverse applications of AI for Text-to-Image Generation,represented by Sora,Midjourney,and Stable Dif fusion,have brought new perspectives to discussions about issues such as training data infringement,copyright confl icts,and the positioning of intellectual property. The intellectual property risks associated with AI for Text-to-Image Generation include potential infringement in the composition of training data,unclear attribution of intellectual property rights,and the expansion of infringement through commercial applications. To ef fectively address these risks,measures such as enhancing the uniqueness and traceability of images,improving the regulation of infringement liability and responsibility,and promoting industry self-regulation and copyright awareness can help foster the compliant,e ffi cient,and sustainable development of AI for Text-to-Image Generation.更多还原