<p>This article develops a threshold-ethical framework for a hypothetical but structurally specifiable class of artificial system-lineages. It proposes the term origin-producing systems for artificial lineages in which systems do not merely produce outputs, but participate in the generation, modification, or authorization of successor systems in ways that may weaken the traceable relation between human design intention, system behavior, and moral accountability. The article does not claim that such systems presently exist in completed form. Rather, it asks whether existing trajectories in autonomous optimization, self-modifying software, automated system design, machine-learning opacity, and artificial agency reveal partial capacities whose convergence would challenge ordinary assumptions about responsibility, governance, and human control. It distinguishes unexpected outputs and ordinary subgoal formation from evaluative drift: a possible shift in the criteria by which future behavior is selected, preserved, or reproduced across iterations. It also clarifies ancestral opacity as a proposed term for opacity across artificial lineages, where the pathway from human intention to later behavior may become practically unreconstructible under conditions of scale, speed, recursive mediation, and documentation loss. The article argues that some systems may require moral evaluation before creation rather than governance only after deployment. The ethics of non-creation is defended conditionally: not as a rejection of artificial intelligence, but as a candidate moral duty for a narrow class of systems that could undermine later accountability, interruption, and control.</p>