If I build an AI that is biased toward creating malware, is it the AI's fault? I say "maybe" because the calculations are performed by the AI, you say "maybe

If I build an AI that is biased toward creating malware, is it the AI's fault? I say "maybe" because the calculations are performed by the AI, you say "maybe not" because of the creators' influence, but it still makes sense to me that we should debug/firewall/shut down that AI.