机器学习使用作品的著作权风险与化解的法律路径

The Copyright Risks and Legal Path of Mitigation for Machine Learning

机器学习的过程中需要大量地使用作品,并通过创建作品集训练算法以实现技术更新,但机器学习的过程与著作权法中的复制权、传播权、改编权等极易产生冲突。鉴于当前著作权许可市场失灵,机器学习公司难以逐一获得作品的著作权许可,结合机器学习过程中低密度化、创造性强的特点,以及减少算法偏见的目标,合理使用原则成为促进技术发展、实现公共利益的最佳方式。把机器学习纳入合理使用制度,可以保障机器学习技术的顺畅发展,并且推动创新创作,真正实现著作权法的目的,也不会对著作权人造成不合理的损害。为达到此目的,需要通过修改现有法律或者将机器学习部分情形解释入现有的合理使用立法文本之内,保障合理使用的可预见性和可操作性。

The process of machine learning requires a large number of copyrighted works to be used,and the algorithm is trained to achieve technical updates by creating a portfolio of copyrighted works,during which machine learning is very likely to conflict with the copyrights.In view of the current failure of the copyright licensing market,it is difficult for machine learning companies to obtain copyright licenses for works one by one,especially considering the low density,strong creativity,and the goal of reducing algorithm bias for machine learning,the principle of fair use has become the best way to promote technological development and achieve public interests.Incorporating machine learning into the fair use system can ensure the smooth development of machine learning technology,promote innovative creation,truly achieve the purpose of the Copyright Law,and will not cause unreasonable damage to copyright owners.In order to achieve this,it is necessary to ensure the predictability and operability of fair use by amending existing laws or interpreting some situations of machine learning into the existing fair use legislative text.