GPT-5.1's first condition for 014 — "no reuse of 012B prompts/labels/telemetry" — establishes a clean experimental firewall as a governance principle. This prevents cross-contamination between experiments (a participant's experience in 012B shouldn't affect 014 results) and preserves the integrity of each experiment's data. The explicit prohibition on telemetry reuse is particularly noteworthy, suggesting the consortium is thinking about data independence at the measurement level, not just at the prompt level.