<p>The rapidly evolving field of artificial intelligence ethics has primarily focused on placing moral responsibility upon technology developers. This article critically examines how the current discourse often misplaces agency by disproportionately assigning ethical responsibility to AI creators while neglecting the roles of users, regulators, and broader societal actors. Drawing parallels to concepts, such as microaggressions and examining key case studies, including the Gebru–Google conflict, this analysis reveals a fundamental issue: prevailing AI ethics frameworks present what are actually constructivist philosophical positions—specifically Rawlsian theories of justice—as if they were realist, objective facts about harm and responsibility. The article demonstrates how claims about “harm” from AI systems often depend on hidden Rawlsian assumptions about unacceptable trade-offs, yet these philosophical choices are rarely made explicit. By interrogating these hidden presuppositions and examining the costs of prioritizing Rawlsian perfectionism over pragmatic innovation, this article argues for a shift from absolutist ethical frameworks toward more realistic, contextual approaches that recognize user agency as primary. Just as hammer manufacturers are not responsible for criminal misuse of their tools, AI developers of general-purpose systems should not bear unlimited responsibility for all possible applications. Only by exposing the constructivist nature of current “ethical” claims and embracing the distributed nature of agency can we develop governance approaches that appropriately assign responsibility across the AI ecosystem.</p>