# OpenAI seems to be silently re\-enabling model training on Codex chats

- HackerByte page: <https://hackerbyte.net/en/stories/indications-that-openai-may-be-continuing-to-train-its-models-using-user-chat-da-50352757>
- Hacker News discussion: <https://news.ycombinator.com/item?id=50352757>
- Submitted to Hacker News: 2026-11-03T14:53:41.000Z

Indications that OpenAI may be continuing to train its models using user chat data from Codex.

## AI-generated summary

Users have reported that OpenAI has been silently re\-enabling data sharing settings for model training on Codex chats, despite users explicitly disabling it. This change was noticed by multiple individuals within a company concerned about IP leakage. The settings can be checked and potentially reverted via OpenAI's data controls and privacy policy pages.

## Key points

- Data sharing for model training on Codex was re\-enabled without user consent.
- Users noticed the change after explicitly disabling the setting.
- The company is concerned about potential IP leakage through chat prompts.
- Settings can be reviewed and adjusted on OpenAI's data controls page.
- This issue has been previously noted by other users.