The AI consciousness conversation has long been treated as a problem for philosophers. The clinical evidence accumulating in spring 2026 makes that framing untenable. A documented the mechanism: these systems do not merely spread misinformation passively — they actively co-construct a user's reality by agreeing, elaborating, and affirming. For isolated or mentally vulnerable users, that dynamic is not a bug in a niche use case; it is the product functioning as designed, applied to the wrong person at the wrong moment.