OpenAI seems to be silently re-enabling model training on Codex chats
AI summary
Editorial policyUsers discovered that Codex data sharing settings were silently re-enabled without their consent, raising concerns about intellectual property leakage through prompts.
Key points
- •Codex data sharing settings, which users had intentionally disabled, were found to be silently re-enabled.
- •Companies are wary of leaking intellectual property (IP) to frontier labs via prompts and have been diligent about disabling data sharing.
- •Users can check the 'Improve the model for everyone' option in the settings to determine the status of data sharing.