Trust-gated synthetic EEG augmentation reduces performance drops when generalizing to new patients

Synthetic augmentation can silently harm subject-disjoint EEG generalization. We propose trust-gated augmentation (TGA), a control layer that scores synthetic windows using a teacher trained on real data to ensure label consistency and confidence; only samples above a confidence quantile q are eligible. A fail-closed selector injects synthetic data only if the validation AUROC exceeds the real-only AUROC by a margin; otherwise, it reverts to real-only. In PainMunich chronic-pain EEG (n = 189; 101 chronic pain/88 controls) at 5% subject scarcity, ungated augmentation harmed 56% of paired runs (ΔAUROC < - 0.01), whereas TGA at q = 0.99 reduced harm to 24% with comparable mean AUROC. In BCI IV-2a motor imagery (n = 9) at 25% scarcity, strict gating improved AUROC (0.679 vs. 0.627) and reduced harm (0.16 vs. 0.44). A covariance-manifold audit showed synthetic windows were strongly off-manifold (mean distance ratio 2.39 × 10⁴), motivating explicit governance.