OpenAI is using humans to read your conversations with ChatGPT: Report
Last Updated: 2026-09-15 13:54:29
ChatGPT is used by millions and a new report claims OpenAI is using humans to read and review these conversations to make its AI better. The details sourced by 404 Media in its report, claims the new information, spotted in a leaked internal document, suggests OpenAI is running this under the condename Project Lily, and reviewing chats done by humans can also include sensitive personal information.
Show AI Summary
Summarized by AI.
OpenAI uses humans to review ChatGPT chats.- Reviews aim to improve AI models (Project Lily).
- Sensitive personal data may be included in reviews.
The concerns of AI models have increased in recent weeks and this new report adds further fuel to the privacy rules of the companies who are running the AI show in the market.
Latest and Popular Mobiles
Reading your chats
This is not the first instance where AI chatbots are allegedly being trained using personal chats. The report says OpenAI is using a different mechanism to use these conversations to train the AI and asking the human reviewers to check for precise methods of the responses, and how it summarises a topic.
Related Articles
Their model also includes giving these AI models a score based on the level of their responses that can range from unacceptable, unusable to hard to help with any improvement of the AI model.
The report further states that OpenAI wants these reviewers to make ChatGPT sound less like a human but their responses should align with their thought process and still match the tone of the person.
Identities masked but…
OpenAI might be getting humans to review those chats but the report claims the company does not reveal the usernames of these people, which is some relief from a privacy point of view. AI models are getting better and even powerful enough to help bad actors build new attacks.
These human reviews need to be disclosed publicly so that people can decide whether they want OpenAI using their private conversations to improve its AI models and make them more feature-rich and natural sounding in the future.