I could write a program today that would insist it is alive and beg for its life. With a decent LLM behind it, I could easily, today, make it convincing. Is that program self aware? I think most of us would argue not.

Star Trek computers have never been AI, in any way an expert would define truly self-aware artificial intelligence. There's no justification about why they neither have nor use AI, but Data was declared as distinct from Trek computers specifically because he had a positronic brain which, in classic Trek terms, somehow made him different from Trek computer programs. Trek programs, and especially holodeck ones, were classified as "simulations." Sometimes, as someone else pointed out, simulations could escape their boundaries and become truly self- aware and arguably truly AI.

There's probably some deep discussion on a Trek board about this subject, but IMO, in the Trek universe Trek engineers and scientists have a more clear understanding and better definitions for AI; and what distinguishes a truly alive, self-aware entity from a merely very good simulation of one. When things like Moriarty come up, they're plot devices to play on our poorly understood definitions and distinctions, to drive a story. Was Vaal sentient? It was hostile, so it probably made no difference, but the characters seemed to easily separate it into the category of "just a machine."

I believe that there's simply some given understanding - probably some basic theory taught at school - that allows characters to make the distinction; some bit of Trek advanced knowledge we haven't yet discovered. Most computers in Trek are not capable of producing true AI, only very convincing simulations, and these are not considered alive, or have rights.