It seems it would be OK if it were limited to being a research assistant and everything produced was rigorously checked. But because AI has no concept of truth, ethics or morals, I would limit it to that alone.

An AI is not capable of not lying, which seems to violate the Ninth commandment. I am not an expert, so interpreting the "against thy neighbor" portion may or may not apply here. I tend to interpret that as "no lies" in any circumstances.

I do realize that because of the way our brains, etc. work, that it's pretty much impossible to not lie, at last a bit, but that is what forgiveness is for. I don't think you can forgive software, since it really should be a correctable issue. Of course, LLMs are so complex as to be uncorrectable.