What we know

Users on Hacker News have noted signs that OpenAI may have quietly re-enabled model training on Codex chats. This speculation arises from observed changes in the user interface and updates to privacy policy language. However, OpenAI has not officially confirmed these changes, and there is no independent verification available.

Why it matters

OpenAI previously allowed users to opt out of having their data used for training, making any shift in data usage policies significant to developers and the AI community. Codex, OpenAI’s code generation model, relies heavily on training data, so changes to how user interactions are handled could impact privacy and model development. These discussions highlight ongoing concerns about transparency and user control over data in AI services.

What is still unknown

It remains unclear whether model training is fully re-enabled or limited to certain interactions. No details have been provided about the accuracy impact, supported programming languages, or any timeline for broader rollout. No independent tests of these changes or their effects on user data privacy have been conducted.