Recently, the data source and copyright issues of generative artificial intelligence models have attracted much attention. Controversies surrounding the copyright of AI model training data are increasingly intensifying, and all parties are seeking effective ways to balance innovation and copyright protection. To solve this problem, some organizations and institutions are actively exploring solutions, such as launching certification labeling programs to ensure the legitimacy of model training data, and promoting relevant legislation to protect the interests of copyright holders.
In response to the issue of whether content generated by artificial intelligence models infringes copyright, the non-profit organization Fairly Trained launched a certification labeling program. The organization has issued a label called "Licensored Model" certification to nine AI companies to ensure that the training data for their models does not infringe copyright. However, controversy continues over generative AI’s use of copyrighted data for model training. At the same time, some bills also require AI companies to disclose the sources from which they obtain training data to protect the rights and interests of copyright holders.
Copyright issues are challenges that must be faced on the road to the development of artificial intelligence. Fairly Trained’s certification program and related legislative attempts have provided useful exploration to solve this problem. In the future, more similar efforts are needed to standardize the acquisition and use of AI model training data to promote the healthy development of artificial intelligence technology and safeguard the legitimate rights and interests of copyright holders.