The good person out of the good treasure of the heart produces good, and the evil person out of evil treasure produces evil; for it is out of the abundance of the heart that the mouth speaks.
Luke 6.45
Ethics: Training Material
Copyright law allows for fair use of copywritten material for review, research, etc. What the AI is doing is not a violation of copyright, but training so that it can produce, not the exact same material, but its own writing. Imagine it as a child reading hundreds (tens of thousands) of books, that material trains the child in how to write well and may effect the "voice" or style of the child. The same thing happens with the AI.
Four guiding principles for fair use, according to Copyright Code:
This is similar to Turnitin and how it trains on thousands of copywritten documents each day. It is not plagiarism, nor violation of copyright, but fair use because it is using that material in a transformative manner; it is not reproducing the material for the sake of profit.
Generative AIs produce transformative works; they do not reproduce works that they were trained on. Yes, certain people were able to "break" a couple of earlier AIs to get to training material, but that is beyond what most people are capable of.
Ethics: Training Material II
Many AI models and tools now train on all materials that are put into them; all prompts and files are used as training material. This is an issue for copyright violation and the issues surrounding privacy. Copyrighted materials can be used by an AI because the AI "transforms" the work into something new. But, there are now lawsuits about owners of copyright not wanting, not giving permission, for their work to be used by an AI. [Personally, I am not a lawyer, but I don't see how this stands up in court. The transformative nature of what an AI does means that it should be perfectly legal to be used in training an AI.]
How this effects the students, is that several of our database providers / vendors are now adding statements in the licensing about the articles they provide; they should not be put into an AI. The database companies are trying to avoid lawsuit by saying they attempted to keep students from placing the document into an AI; keeping the AI from training on the article.
[Snarky thoughts - none of the following can control what a student does with a PDF: vendor, librarian, nor professor. Also, every piece of software used to view a downloaded document now has an AI built-in; Adobe Reader, MS Word, etc. There is a possibility that the vendors are also desiring that the article is only viewed online, in the vendors viewer, which has its own AI built-in (JSTOR and others)]