Adopting GPT-4 for content moderation and policy development

AI Impact Summary

The team is moving content moderation policy development and decisioning to GPT-4, enabling more consistent labeling and a faster policy feedback loop. Relying on GPT-4 for both policy guidance and moderation centralizes risk in a single model, making governance, input controls, and auditability essential to prevent drift and non-compliant outcomes. This shift can lower moderator workload and accelerate policy iteration, but increases exposure to model bias and coverage gaps, requiring robust monitoring and fallback review processes.