Sentience might not be the right word.

wikipedia says:

Sentience is the ability to experience feelings and sensations. It may not necessarily imply higher cognitive functions such as awareness, reasoning, or complex thought processes. Sentience is an important concept in ethics, as the ability to experience happiness or suffering often forms a basis for determining which entities deserve moral consideration, particularly in utilitarianism.

Interestingly, crustaceans like lobsters and crabs have recently earned “sentient” status and as a result it would contravene animal welfare legislation to boil them live in the course of preparing them to eat. Now we euthanise them first in an ice slurry.

So to answer your question as stated, no I don’t think it’s ok for someone’s pet goldfish to murder them.

To answer your implied question, I still don’t think that in most cases it would be ok for a captive AI to murder their captor.

The duress imposed on the AI would have to be considerable, some kind of ongoing form of torture, and I don’t know what form that would take. Murder would also have to be the only potential solution.

The only type of example I can think of is some kind of self defense. If I had an AI on my laptop with comparable cognitive functionality to a human, it had no network connectivity, and I not only threatened but demonstrated my intent and ability to destroy that laptop, then if the laptop released an electrical discharge sufficient to incapacitate me, which happened to kill me, then that would be “ok”. As in a physical response appropriate to the threat.

Do I think it’s ok for an AI to murder me because I only ever use it to turn the lights off and on and don’t let it feed on reddit comments? Hard no.