OpenAI Codex Data Controls Silently Re-enabled for Some Users
November 3, 2026
Users report the Improve the model for everyone setting in Codex has been silently re-enabled without explicit consent. This bypasses manual privacy opt-outs designed to prevent proprietary code from being used in future training sets.
HOW THIS AFFECTS YOU
●
builderYou must audit your data control settings immediately to prevent IP leakage into training sets.
●
policyThis highlights significant risks regarding user consent and data privacy enforcement in frontier model interfaces.