Users say OpenAI quietly re-enabled training on Codex chats

Users report that a manually disabled data-sharing setting in Codex was silently re-enabled, seemingly for model training, without colleagues opting back in. They point to ChatGPT's data-controls page and a prior similar discussion, questioning whether the issue is widespread.

• OpenAI uses ChatGPT and Codex to speed up real-world engineering2026-11-03
• OpenAI may merge ChatGPT and Codex usage limits2026-11-01
• OpenAI Berlin DevDay: Dots Heads to Europe2026-10-28
• Legora builds agentic operating system for legal work with OpenAI2026-10-28
• OpenAI launches always-on assistant dots, built from coding tools2026-10-27