随着科技的进步，信息网络已经成为现代人类生活的基础，随之发展而起的各项技术也都发挥着多面且不可替代的作用。其中，人工智能是新一轮科技革命的核心驱动力量。生成式人工智能的出现标志着人工智能正式进入2.0时代，这种产出模式在大大提高创作效率的同时也丰富了创作的多样性，但随之引发的一系列法律问题也引人深思。

生成式人工智能运行的根本逻辑是通过对大量作品数据进行采集、解构，经过一系列深度机器学习，对处理后的作品数据进行再次搭建，从而形成具有独创性的新内容。这其中涉及到数据抓取、解码重组、再次输出等行为，对著作权中的复制权、改编权、汇编权、传播权等权利均有不同程度的触及。分析生成式人工智能的运行操作是否构成版权法中相应权利的行为，行为是否经过原作品著作权人的授权许可等问题是判断是否构成侵权的重要标准。显然，基于生成式人工智能对数据大量需求及自行收集的特点，其对数据的取用构成版权法中的复制。而其大量的自行复制行为要一一得到作品数据权利人的许可是很难成立的，故而在当下学界的普遍认识中，认为此举会构成对原作品复制权的侵犯。而对于改编权、汇编权等在机器学习阶段可能触及的权利来说，虽然在操作形式上与版权法所规定的相应行为有些许重合，但因生成式人工智能的数据分析仅发生于算法逻辑的框架内部，属于一种生成过程中的技术性行为，并不构成对原作品相应权利的侵犯，故而不宜划入侵权的范围。而在内容输出阶段，对外传输数据生成结果的行为属于版权法中的传播，结合当前司法实践中普遍适用的“服务器标准”，可判定其构成信息网络传播权侵权。此外，对于生成式人工智能的生成内容本身是否构成对作品数据复制权的侵犯，则需要根据“接触+实质性相似”的标准具体情况具体分析。

从整体来看，人类对生成式人工智能更为广泛的运用已是当下大势所趋，但技术的更迭并非一朝一夕之事，需要人力物力长久的投入。在相关立法仍处推进中时，为了使技术更好的服务于人类创作，实现该项技术的市场价值最大化，则要使被认定为侵权的情形脱离违法范畴，根据当下立法情况找到适宜路径，也即对生成式人工智能著作权侵权责任的豁免展开思考，分别判断其对于合理使用制度与法定许可制度的适用可能性。最后，结合生成式人工智能发展趋势，对其著作权侵权问题的规制提出更具针对性的建议。

With the progress of science and technology, the information network has become the foundation of modern human life, and the technologies that have developed along with it have also played a multi-faceted and irreplaceable role. Among them, artificial intelligence is the core driving force of the new round of scientific and technological revolution, the emergence of generative artificial intelligence marks the 1.0 era of artificial intelligence officially entered the 2.0 era, this output mode greatly improve the efficiency of creation at the same time also enriches the diversity of creation. However, a series of legal issues arising from it are also thought-provoking.

The fundamental logic of generative AI operation is to collect and deconstruct a large amount of work data, and after a series of deep machine learning, the processed work data will be built again, so as to form new content with originality. This involves acts such as data capture, decoding and reconstruction, and re-exporting, which touch on the rights of reproduction, adaptation, compilation and dissemination in copyright to varying degrees. Analysing whether the operation of generative AI constitutes an act of the corresponding right in copyright law, and whether the act has been authorised by the copyright owner of the original work is an important criterion for judging whether it constitutes an infringement. Obviously, based on the characteristics of generative AI's massive demand for data and self-collection, its access to data constitutes copying in copyright law. And its large number of self-copying behaviour to get the permission of the work data right holder is very difficult to establish, so in the current academic community's general understanding, it will constitute an infringement of the right of reproduction. As for the right of adaptation, compilation and other rights that may be touched in the stage of machine learning, although there is a slight overlap between the form of operation and the corresponding acts stipulated in the copyright law, the data analysis of generative artificial intelligence only occurs within the framework of algorithmic logic, which is a kind of technical behaviour in the process of generating, and does not constitute an infringement of the corresponding rights of the original work, and therefore is not suitable to be classified into the scope of infringement. In the content output stage, the act of transmitting the data generation results to the outside world belongs to the dissemination of copyright law, combined with the "server standard" generally applicable in the current judicial practice, can be judged to constitute the infringement of the right to disseminate information network. In addition, whether the generated content of generative AI itself constitutes an infringement of the right to reproduce the data of the work needs to be analysed on a case-by-case basis according to the standard of "contact + substantial similarity".

From an overall perspective, the wider use of generative AI by human beings is a general trend at present, but the change of technology is not an overnight matter, and requires long-term investment of human and material resources. When the relevant legislation is still in progress, in order to make the technology better serve human creativity and maximise the market value of the technology, it is necessary to make the infringement situation out of the scope of the law, and to find a suitable path according to the current legislation, i.e., the exemption of generative AI copyright infringement liability to think about, and to judge the applicability of the fair use system and the statutory licensing system respectively. Finally, in light of the development trend of generative AI, it will put forward more targeted suggestions for the regulation of its copyright infringement.