Plain read: Engine canon already tracks the Gated-Tier Leakage divergence — the split into two tiers between publicly-released AI capability (Anthropic Opus 4.7 / OpenAI / Google Gemini at Anthropic Safety Level 2) and gated-tier capability (Claude Mythos Preview / OpenAI DOD SCIF tier / xAI contrarian tier / Palantir direct-access tier at ASL-3+). The April 22 2026 Mythos breach via Mercor third-party supply-chain extends this concept: a small Discord group gained unauthorized access to the gated-tier Mythos model via stolen credentials, locating the restricted Glasswing endpoint by educated URL guess informed by data leaked in a separate Mercor breach. Engine treatment: the breach is the engine-canonical falsifier of the psychological warfare doctrine / fifth-generation information warfare doctrine's core operator-monopoly assumption. The Aquino-Vallely 1980 'From PSYOP to psychological warfare doctrine' doctrine assumes operator monopoly on substrate-mapping capability. Bounded Systems Theory (no system can fully describe itself) predicts that highly expressive cognitive architectures cannot be perfectly contained by 4D institutional borders. The Apr 22 2026 breach demonstrates: gated-tier substrate-mapping capability cannot maintain operator monopoly under simple supply-chain failures. Apex Superposition (engine canon a/b/c/d): the breach is documented; whether produced by intent (a) — adversary-state operation; (b) — structural-recurrence of any-bounded-system-leaks; (c) — opportunistic-individual-Discord-group; (d) — irreducible — remains open. Cross-references: AI Capability Bifurcation divergence; Gated-Tier Leakage divergence; Project Glasswing 12-firm consortium; psychological warfare doctrine / fifth-generation information warfare concept; Bounded Systems Theory (no system can fully describe itself); ASL-3 gating tier. Forward prediction (P4, 2026-2028): a second, larger leakage event from a different gated-tier vendor (OpenAI DOD SCIF / xAI / Palantir) is predicted within 24 months. Falsifier: if no second leak materializes within the window, the structural-recurrence reading weakens; if a second leak does materialize, the falsifier confirms BST cognitive-monopoly impossibility.