An Ars Technica reader has raised concerns about potential data leakage from ChatGPT, an AI-powered chatbot developed by OpenAI.

Screenshots submitted to the publication on Monday suggest the unauthorised disclosure of sensitive information from private conversations, including login credentials, personal details, and other confidential information.

Following the article’s release, OpenAI has said currently investigating the matter.

What was Leaked and how?
The submitted screenshots provide insight into a troubling breach of privacy within ChatGPT. Two instances, in particular, reveal compromised login credentials and personal details. One screenshot captures a conversation between an employee and the AI chatbot, expressing frustration over technical issues with a pharmacy prescription drug portal. The conversation inadvertently led to the disclosure of usernames and passwords.

The leaked conversations reveal a pattern of data exposure extending beyond login credentials. Identifying information, such as the application’s name and store number associated with the reported problem, are potentially catastrophic with regards to security to data protection.

More Questions Raised about ChatGPT
The Ars Technica report isn’t the first time OpenAI’s LLMs have raised concerns. ChatGPT came under fire late last year in Poland for potentially breaching GDPR, the EU’s data protection and digital privacy regulations.