OpenAI silently re-enabling model training on Codex chats

A user on Hacker News reported that the data‑sharing setting for OpenAI’s Codex chat had been silently re‑enabled. The user and coworkers, who had previously disabled the “Improve the model for everyone” option to protect intellectual property, noticed the setting had changed without their action. The setting can be checked at https://chatgpt.com/settings/data-controls, and the user notes that the option can also be turned off via https://privacy.openai.com/policies?Modal=take-control.

The issue was flagged earlier on Hacker News (item 49643556) and has since generated a discussion (item 50352757) with 265 points and 84 comments. The post highlights concerns about unintended data sharing in OpenAI’s code‑generation models.