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

What happened: A user reports that the data sharing setting in Codex was silently re-enabled for them and coworkers, despite the team having disabled it to avoid leaking IP through prompts. The user asks whether the issue is isolated or widespread and links to a prior discussion. Why it matters: The report raises concern among teams that restrict prompt data sharing with frontier labs, since users say they did not knowingly opt back in. The poster points to a privacy settings page where data sharing can be turned off. Key numbers: 17 points, 1 comment.

A user reports that the data sharing setting in Codex was silently re-enabled for them and coworkers, despite the team having disabled it to avoid leaking IP through prompts. The user asks whether the issue is isolated or widespread and links to a prior discussion.

The report raises concern among teams that restrict prompt data sharing with frontier labs, since users say they did not knowingly opt back in. The poster points to a privacy settings page where data sharing can be turned off. Key numbers: 17 points, 1 comment.