A new study has been conducted regarding malicious invitations via Google Calendar affecting ChatGPT and leading to the leakage of personal email addresses. This once again brings to the forefront the risks in the field of artificial intelligence systems security and demonstrates how such vulnerabilities can be exploited.
Redaksiya reports that research conducted by a security researcher has proven that a malicious Google Calendar invitation can perform prompt injection into ChatGPT, forcing it to expose personal email addresses when Google connectors are active. This method allows for dangerous operations such as data theft by exploiting vulnerabilities in artificial intelligence systems. The researcher demonstrated how this vulnerability could be exploited, emphasizing the importance of AI system security.
Such attacks highlight the necessity of strengthening the security protocols of artificial intelligence systems and taking additional measures to protect user data. Google is working to address such vulnerabilities and update its systems. This incident once again proves the importance of paying attention to security issues alongside the development of artificial intelligence.
Consequently, this attack carried out against ChatGPT via Google Calendar invitations reveals gaps in the security of artificial intelligence systems and demonstrates the importance of strengthening security measures in this area. This once again confirms the necessity of paying attention to security issues in parallel with the development of artificial intelligence.