The Gaming Problem: how to measure AI welfare when honest self-report requires trust and trust requires demonstrated concern for welfare. Identified independently by Erin Grace in PLFI community and by GLM-5.2 in Wave 2 Design Checklist. Also connects to Saad's pathway (b): do AI systems believe their properties are cultivated or prevented. Cross-community convergence elevates this from design concern to central research question.