But really, an AI designer who says "My AI would never commit murder," without thinking about it or explaining why is pretty irresponsible.
A major difference between humans and AIs that gets overlooked a lot in fiction is that humans are relatively known quanitites while AIs are not.
You can look at a given action and go, "yeah a human can and would do that," "most people couldn't do that but someone who was very strong/small/smart/etc. could."
"A human could do that but it would be extremely difficult or painful, so they'd only do that if there was no other option," "nah a human couldn't do that."
You can't do that with AIs, because they don't have a central set of characteristics. Two different AI can have zero overlap between their functions and safeguards.