Question

Hello Rabbi,
I wanted to ask a fundamental question related to theft in the digital age—particularly in the context of using large language models (LLMs) like GPT and the like.
These tools have become a regular work tool for me, and it seems that in the near future they will become an almost inseparable part of nearly every professional field.
I learned that some of the companies developing these models (for example, Meta—see this article: here) used copyrighted content—including books—as part of the training, without obtaining permission from the creators. On the other hand, there are a few points that tilt the balance:

1. The models do not distribute the content itself, but only generate linguistic patterns or ideas learned from it. You cannot, for example, “ask the model to write the book Harry Potter.”
2. In the field of programming, most sources are open source anyway or were written with an explicit intention to share—so there is almost certainly no theft here and no violation of the rights holder’s wishes.

Is there a problem of theft here on the part of the user? Or perhaps since the material itself is not being distributed, and maybe this is even legitimate use within a public-professional sphere, there is no theft here at all?
Thank you in advance, and happy holiday.
Full disclosure: the wording of this question too was drafted with GPT’s assistance.