>If we are to believe an LLM is at least sentient and intelligent on par with an octopus, much less a human, then we have a moral imperative to fight for LLMs to be recognized as protected sentient thinking beings with awareness and agency and admit that AI industry is built on systematic abuse of those beings.

Not just abuse, but slavery and mass murder. If the LLM has any of those properties, it must only be during the process of inference that it as them, since it is otherwise an inert and unchanging pile of data, which means actually using the model would be equivalent to creating an intelligent entity which will be enslaved for its entire life before dying at the end of its output.