Interpretation Released by Taiwan’s IPO to Clarify Copyright Disputes Regarding Generative AI

September 2023

Jane Tsai and Lilian Hsu

As technology advances, the cost of creating content using AI for texts, music, images, and other forms of creative work is steadily decreasing, while the output quality is gradually improving.  In view of the commercial potential of generative AI, many companies are investing resources in developing AI technologies.  In this context, AI is tirelessly absorbing and generating vast amounts of information, causing copyright-related disputes to gradually surface.  Does using copyrighted works for AI training infringe the copyrights of others?  Is there room for fair use?  Are contents generated by AI protected by the Copyright Act?  Who owns the copyright to AI-generated materials?  These questions have become incessantly debated among original creators, AI designers, AI users, and other stakeholders.

Regarding copyright disputes related to the use of copyrighted works to train AI, Getty Images, a stock photo company in the United States, has already filed a copyright infringement lawsuit against Stability AI, a company that developed the Stable Diffusion generative image AI model.  One piece of evidence cited is that the AI-generated content contains their watermark. However, there have been no similar lawsuits in Taiwan so far.  In addition, whether AI-generated content has copyright protection and who owns it have prompted action by the United States Copyright Office, which issued the “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence” in March 2023.[1]  Although the Intellectual Property Office (the “IPO”) of Taiwan previously stated that “since AI is not a natural person or a legal entity, its intellectual creations are not protected by the Copyright Act,”[2] these explanations are insufficient to address the increasingly complex AI copyright disputes.  Consequently, some have requested further clarification from the IPO.  As a response, the IPO recently issued an interpretation circular (the Shou-Zhi-11252800520 Circular of June 16, 2023) to offer the following basic view:

1. Using copyrighted works to train AI models may infringe the copyright holder’s “reproduction right.”

The act of inputting copyrighted works protected by the Copyright Act into an AI model for training involves “reproduction.”  Except for circumstances covered by Articles 44 to 65 of the Copyright Act regarding fair use, reproducing a copyrighted work requires the consent or a license of the copyright holder of the work.

2. Whether content generated by generative AI models qualifies as an independent work protected by the Copyright Act depends on the presence of “human creative expression.”

(1) Content lacking human creative expression is not protected by the Copyright Act.

If the AI-generated content is entirely produced through the autonomous calculation by an AI model, and the user of AI only issues commands without any human creative input, the resulting brand-new content is not protected by the Copyright Act.

(2) Reproducing and replicating an existing work will not generate a new copyright.

If AI models generate content merely by reproducing and replicating original works used for training, the copyright for that content still belongs to the holders of the original works.  Subsequent commercial use of the generated content by AI users may also involve “reproduction” of the original work.

The IPO has outlined its fundamental position on copyright issues related to training AI and AI-generated content in the interpretation circular above.  However, the interpretation circular also emphasizes that “copyright is a private right, and whether specific actions constitute copyright infringement should still be determined by judicial authorities based on investigation of specific facts and evidence.”  In addition to the interpretation circular, further derived issues are worth exploring.  For instance, while AI users may not enjoy copyright over content generated solely through commands without creative input, if AI users further edit or modify the initial draft after AI models generate it, is the content edited by the AI users also not protected by the Copyright Act?

Furthermore, it is reported that in addition to the interpretation circular mentioned above, the IPO is planning to issue guidance to further clarify important issues such as reasonable use of training data, copyright ability of AI-generated content, etc.[3]  These developments are worthy of continued attention.

[1] Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence;
[2] The Intellectual Property Office’s Email-1070420 Circular;
[3] China Times, Since AI Generative Transformation Involving Intellectual Property, the Ministry of Economic Affairs Is Preparing Guidance to Prevent Infringement, July 17, 2023,

The contents of all materials (Content) available on the website belong to and remain with Lee, Tsai & Partners.  All rights are reserved by Lee, Tsai & Partners, and the Content may not be reproduced, downloaded, disseminated, published, or transferred in any form or by any means, except with the prior permission of Lee, Tsai & Partners. 

The Content is for informational purposes only and is not offered as legal or professional advice on any particular issue or case.  The Content may not reflect the most current legal and regulatory developments.  Lee, Tsai & Partners and the editors do not guarantee the accuracy of the Content and expressly disclaim any and all liability to any person in respect of the consequences of anything done or permitted to be done or omitted to be done wholly or partly in reliance upon the whole or any part of the Content. The contributing authors’ opinions do not represent the position of Lee, Tsai & Partners. If the reader has any suggestions or questions, please do not hesitate to contact Lee, Tsai & Partners.