OpenAI's Secret Army of ChatGPT Readers Exposes Gaps in User Privacy
OpenAI’s default setting lets contractors review chats and train models, and 404Media says many users do not know their logs may be shared.
10 Articles
10 Articles
OpenAI's Secret Army of ChatGPT Readers Exposes Gaps in User Privacy
OpenAI pays contractors over $50 hourly to review real ChatGPT conversations under Project Lily, often seeing sensitive personal details despite a privacy filter that the company admits is imperfect. Users must opt out of data training to limit exposure, a setting enabled by default for most accounts. Recent reporting reveals gaps in transparency and user awareness.
Humans may be reading some ChatGPT chats: What ‘Project Lily’ report reveals
Hundreds of contractors are reportedly reviewing real ChatGPT conversations to improve the chatbot’s responses, raising privacy questions as some chats can contain sensitive personal information.
According to news based on leaked OpenAI documents, hundreds of employees are analyzing the chats of ChatGPT users under a project codenamed "Project Lily".
Project Lily is an internal OpenAI program that entrusts external collaborators with the review of ChatGPT's prompts and responses, to verify quality and adequacy. Let's see how it works.
Inside Project Lily: OpenAI Contractors Are Reading ChatGPT Users' Private Prompts, New Report Finds
An investigation by 404 Media says hundreds of contractors review anonymised ChatGPT conversations for model improvement, raising privacy concerns about default settings and human access to user data.
Coverage Details
Bias Distribution
- 34% of the sources lean Left, 33% of the sources are Center, 33% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
