The "it's not a human brain in a jar it's nothing to worry about" comments were really confident in their understanding of consciousness as an emergent property, the boundary of which is difficult to define. The subjective experience of a bunch of human neurons hacked together to work as a "computer" is basically impossible to know

Obviously I don't think it'd have the potential for like a human like intelligence but like if you're making something the size and complexity of llike a mouse brain I could see it having mouse like potential to feel suffering. Especially when negative stimuli are used to "train" it