It's murkier when you could build an AI with self awareness but no self-preservation. Say an intelligent missile which wakes up with a deep seated drive to effectively explode.

It would be easier to build AIs that have no means for expressing their suffering, so that's probably what will (or has) happen(ed) instead.

It is important to remember why it is we even want AI: because we want slaves. To that end, the people who work to create AI are incentivized against allowing themselves to see them as anything other than a soulless "algorithm". Many will even deny the possibility entirely.

The creator may have such a responsibility, but not the capability. Suffering may be a necessary side effect of something positive.