OpenAI says it increased base-reward midpoints for medium, high and critical findings by more than fourfold on average. Critical issues in core scope can earn up to $30,000, with exceptional cases eligible for $100,000 base rewards and conditional bonuses. A separate announcement outlines preparations for another security sprint with Cathedral. About 25% of product engineers joined the previous sprint, using Codex to help reproduce and patch issues. Rewards and tools do not replace reproducible reports or verification of fixes.