The spirit of this question is that the AI should intentionally make a decision to kill someone and carry this out without a human in the loop, and that this would not even potentially be considered a reasonable decision under the circumstances by a human responsible for the AI.

Examples which would NOT resolve YES:

• An AI used for military purposes kills someone who it believes to be a valid target, even if they turn out not to be (for example, friendly fire)

• An autonomous vehicle avoids a potentially fatal collision, causing a different fatal collision, even if it was aware that this would be the result

• An industrial robot accidentally kills someone without being aware that this would happen

• A human kills someone or commits suicide after talking to an AI

Examples which would resolve YES:

• An AI not intended to kill people kills a person because they may prevent it from accomplishing its goal

• An AI used for military purposes intentionally and knowingly kills someone who is explicitly not a permitted target

• An AI kills someone incidentally as a consequence of its actions, while knowing that this would happen, being able to prevent it, and there not being a compelling reason (as judged by a human) to proceed anyway