CSENI-S v1.1 is a theoretical and architectural continuation of the CSENI / NIESC framework for non-invertible external supervision of advanced AI systems. This work does NOT claim an empirically validated architecture or a completed proof of containment. It proposes a multi-level sovereign containment architecture with explicit state machines, operational contracts, deployment profiles (MXC/ORC/ZSC), human habitability metrics, and falsifiable hypotheses. Thesis: under auditable assumptions, a stratified architecture can verifiably increase the cost, time, coordination, and detection probability of silent evasion attempts compared to non-stratified equivalents. Includes: axioms, startup assumptions, adversary tiers (T1-T3), defender cost model, RUN/HOLD/SHUTDOWN/SAFE_DEGRADED state machine, dual baseline governance, graded critical memory, and L0/L1/L2 falsification protocol. Spanish version included as supplementary file. Related preprint: CSENI / NIESC - Control Supervisado Externo No InvertibleWebsite: https://josemrivera.com