GCCL v0.1 defines a tiered evaluation and certification framework for AI language systems deployed in organizational contexts with operational risk exposure (R3 and above). The specification establishes: task-generator-based domain evaluation to prevent benchmark overfitting; a dual-track drift detection regime (configuration drift τ and Behavioral Drift Index BDI); a Constraint-Boundary Variance Monitor (CBVM) as prospective early-warning signal; a formal five-level autonomy scale (A0–A4); and two compliance modes (Core and Extended) calibrated to organizational implementation capacity. GCCL v0.1 is designed to be compatible with ISO/IEC 42001:2023, EU AI Act (2024/1689), and enterprise SLA structures. Key parameters requiring empirical field calibration are explicitly identified as open parameters. This record is part of a three-layer governance architecture:- GCCL (system-level capability certification)- Decision Identity Protocol / DIP-CORE-1.0 (decision-level identity and re-legitimation)- Output-Only Diagnostics (doi:10.5281/zenodo.18361523) (observation-level structural drift detection) Status: Working Draft (WD). No part of this draft shall be cited as a finalized standard.