A German court ruled that OpenAI's training of ChatGPT on song lyrics amounts to copyright infringement.
According to dpa and Reuters on the 11th local time, the German music rights association (GEMA) said it won a lawsuit it filed, claiming OpenAI used the lyrics of nine German songs without permission. The court ordered OpenAI not to store the disputed lyrics or output them in responses. It also ordered OpenAI to disclose the records of its use of the lyrics and the revenue earned from it, along with damages.
OpenAI argued that training using song lyrics is a "sequential analysis, a combination of iterative probabilities," saying the association misunderstood how ChatGPT works. But the court accepted the association's claim that using song lyrics constitutes unauthorized reproduction and performance. It said the lyrics were stored without permission and pulled out as is when needed.
The association said OpenAI trained ChatGPT on German song lyrics without entering into a license agreement or paying fees, and brought suit citing nine hit songs, including Reinhard Mey's "Über den Wolken" (Above the clouds).
OpenAI signaled it will appeal. An OpenAI representative said, "We do not agree with the ruling," and added, "We are considering our next steps." The representative emphasized, "This decision concerns certain song lyrics and does not affect the millions of individuals, corporations, and developers in Germany who use our technology every day."
OpenAI, Google, and other Generative AI developers are facing lawsuits around the world for allegedly using copyrighted content and news articles without authorization. Developers argue that AI data training falls under "fair use" permitted by the Copyright Act, unlike unauthorized copying.