ChatGPT Allegedly Leaks Passwords from Private User Chats
According to a recent report by ArsTechnica, the popular chatbot ChatGPT has been found to publish private conversations, including login credentials and other personal data belonging to unrelated users. The issue was brought to light by ArsTechnica reader Chase Whiteside, who provided seven screenshots, two of which contained several pairs of usernames and passwords used to order prescription medications from a pharmacy portal. An employee using the chatbot appeared to be troubleshooting issues with the portal. The leak included the name of the application where the employee was resolving problems and the store number where the issue occurred. One screenshot showed the entire chat conversation.
The reader received this information while using ChatGPT for an unrelated request—he had asked the chatbot to come up with creative names for colors in a palette. Afterward, he noticed elements from several other users’ conversations in his chat history.
These included the title of a presentation someone was working on, details of an unpublished research proposal, and a script using the PHP programming language. One of the conversations about the prescription portal mentioned the year 2020, while other chats did not include any dates.
Previous Incidents and Company Response
In March of last year, OpenAI temporarily disabled ChatGPT after a bug caused chat titles from some users’ histories to be shown to others. In April, the company allowed users to disable chat history in their privacy settings.
In June, OpenAI faced a lawsuit over the alleged theft of users’ personal data. The plaintiffs claimed the company violated data protection laws and unlawfully collected information from the internet, including personal data from individuals who had not consented to its processing.
In November, researchers published a paper describing how they used prompts to get ChatGPT to reveal email addresses, phone and fax numbers, physical addresses, and other personal data that had been included in the materials used to train the large language model.
Many companies, including Apple, have restricted their employees’ use of ChatGPT and other chatbots.
An OpenAI spokesperson stated that the company is investigating the report.