July 2023
 

Credentials for over 101,000 ChatGPT user accounts identified on the dark web
 

Global cybersecurity firm Group-IB has found logs of information-stealing malware being traded on the dark web between June 2022 and May 2023 and identified over 101,000 devices infected with this malware. The highest number of stolen credentials comes from the Asia-Pacific region, accounting for around 40% of the total.

By default, all inputs and AI-generated responses are stored by ChatGPT, which means that any potentially sensitive information users have entrusted to the platform may be at risk. This could be of particular concern for researchers who may have used ChatGPT to help with preparing their research manuscripts. Indeed, Samsung employees inadvertently leaked sensitive company data by sharing it with the chatbot, in what is known as a conversational AI leak.

Researchers should be aware of potential data security risks when using any generative AI, not just ChatGPT.