On Legal Risks of Using ChatGPT for Processing Lockits
Alexander Nasonov https://www.linkedin.com/in/alexander-nasonov/Working with massive multilingual lockits in gamedev projects is a costly process that can be optimized by using such tools as ChatGPT.
Some time ago I had to check the concerns that using ChatGPT for translating, processing or editing localization strings may lead to legal risks. While trying to dig a little bit into the issue I encountered this article that covers a more general topic: The Legal Risks to Using ChatGPT (https://lnkd.in/dEC92Vn9) by American lawyer Leslie A. Farber.
Here are the extracts from the article that I found essential:
“As a language model, ChatGPT is trained on a large dataset of text. Although it generates responses in a human-like way, the output is not considered an original creation of the human mind. Therefore, it is not protected by copyright law and can be used freely.”
“The rights of ownership of AI-generated works will be an issue that continues to evolve along with regulatory changes and case law, but one thing is clear: the content will not be owned by the AI itself. Because AI cannot be recognized as an author and does not have a legal personality, U.S. law prohibits it from owning a copyright.”
The article states that in general the biggest legal risk of working with modern AI-based solutions lies in the field on unintentional plagiarism or unintentional unlicensed usage of previously copyrighted materials:
“For example, if a response includes a quote or image from a copyrighted book, the quote or image is protected by copyright. The same is true if the AI-generated content includes a trademarked brand name or logo. The person using ChatGPT to generate content that infringes on existing copyrights could be held liable for infringement. In order to avoid infringing on the copyright or trademark, the person or entity using the quote, brand name, or logo would need permission from the holder or a license to use the content in a specific way.”
If we go back to the more narrow field of translating, editing and processing the texts of a computer game we should admit that when using ChatGPT or similar instruments in localization work one does not touch any other copyrighted materials other than those that are included into the original game itself. So the whole essence of the job implies that one cannot do any unintentional plagiarism or unintentional unlicensed usage of previously copyrighted materials.
Thus I conclude that using ChatGPT as a tool to do lockit editing is not more risky then using Google Translate or DeepL (which also uses AI based tech) for game texts translation purposes.