Cybersecurity risks regarding ChatGPT account theft and data recovery

Security experts and researchers have highlighted significant privacy risks associated with AI assistants like ChatGPT. According to ESET Latinoamérica, unauthorized access to a ChatGPT account can expose sensitive information, including professional data, client details, medical records, and private projects. Attackers may use this information for identity theft, phishing, or to compromise connected services such as email and document integrations.

Common methods for account compromise include phishing, the reuse of leaked passwords, and malware on compromised devices. To mitigate these risks, experts recommend using robust, unique passwords and monitoring active sessions.

In a separate development, researchers from Georgia Tech have demonstrated that deleted AI conversations can sometimes be recovered from mobile device memory. Using a forensic system called ORISA, the team successfully reconstructed parts of deleted sessions from local AI assistants running on Android devices. The study shows that even after a user deletes a chat through an application interface, fragments of the conversation may remain in the device's working memory.