Content creators’ rights are violated when users of AI services like ChatGPT customize chatbots by feeding them protected content.
When users of AI services such as ChatGPT set out to customize their own chatbots, it is rarely with copyright in mind. Together with Forlaget Praxis and Danske Forlag, we have uncovered that users upload copyrighted content to their own chatbots on AI services. This week, Politiken and the American tech magazine Wired reported this.
These chatbots specialize in reproducing content from the works it is fed with, which is also made available to others, e.g. via OpenAI’s GPT Store, or by direct sharing with friends and classmates. At the same time, the AI services automatically gain permission to train their future AI models on the works in question, unless the user actively opts out.
Read Politiken’s article: Tyveri ved højlys dag med ChatGPT vækker bekymring: Når de bliver taget i det, bliver de ofte flove (Danish)
AI services should protect against infringement
Through searches of chatbots published by users in the GPT Store and elsewhere, it is clear that AI services are full of chatbots to which users have uploaded copyrighted works without permission from the copyright holders. It is largely students and teachers who customize chatbots on material from the curriculum, which are then able to answer questions and reproduce passages from the works. On behalf of Praxis, the Rights Alliance has succeeded in removing chatbots that infringe the publisher’s rights from ChatGPT, after contacting OpenAI. However, it is only possible to take action in cases where infringing chatbots have been made publicly available by the user.
Thomas Heldrup, Head of Content Protection and Enforcement at the Rights Alliance, says:
“When users customize their own chatbots, they do not receive any information from the AI services to make them aware that they must comply with copyright and they do not have to confirm to the AI service that they own the rights to the material they upload. At the same time, rights holders are left with few options for tracing and thus enforcing their rights. We therefore need AI services to provide much better search tools for rights holders, and to take greater responsibility for guiding users to legal use.”
Read Wired’s article: OpenAI’s GPT Store Is Triggering Copyright Complaints