AI & Enterprise
OpenAI faces scrutiny over project using human reviewers to assess real ChatGPT chats
OpenAI is reported to be running an internal project in which human reviewers read and assess real ChatGPT conversations to improve response quality. Documents cited by 404 Media say hundreds of contractors evaluate chats and the model’s answers, sometimes with full conversation context and a user memory summary. OpenAI uses a privacy filter to mask identifiers, but it acknowledges some information can be missed. Users can disable model improvement for new chats.
