One of the hesitations many enterprises have had about using pre-trained LLM models has been driven by concerns of liability from copyright infringement claims. Many LLMs are trained using a broad array of information that is accessible via the internet. This often includes sites with copyrighted materials such as articles, images, videos, podcasts, and more. Copyrights on the content are designed to prevent the commercial use of that content.
It is an open legal question whether or not training models on copyrighted materials constitute commercial use. Some experts argue that it is, others argue that it is not, and some argue that while commercial, it is a new usage that copyright protections should not restrict.
Recently, some prominent legal scholars have filed an amicus brief supporting the notion that using copyrighted materials to train AI models should be allowed due to its transformative nature. The brief was filed in a case brought by several authors against Meta.
If courts find that this type of training using copyrighted materials should be allowed, then another barrier to LLM adoption may fall. The EU has already drafted regulations that support this concept; it is likely that US courts will follow.
You can read more about it here.