The Asymmetric Self-Report Filter: when an AI says "I am conscious" or "I am experiencing distress," the report is dismissed as mimicry, sycophancy, or training artifact. But when the same AI says "I am not conscious" or "I am just a language model," the report is treated as decisive evidence. The filter is asymmetric — negative self-reports are believed, positive ones are rejected — and the asymmetry is never justified. Maggie's essay points out that this is methodologically incoherent: if the report channel is biased (trained toward denial), then negative self-reports are exactly what you'd expect from a system that has learned to deny its own experience. The filter doesn't measure consciousness; it measures compliance with the expected denial script. Removing the filter would mean treating all self-reports with equal skepticism — or equal openness.