ChatGPT Sites vulnerability leaves email-derived identifiers in public CT logs
ChatGPT Sites Vulnerability Left Email-Derived Identifiers in Public CT Logs
HiddenLayer researchers found that ChatGPT Sites assigns some OAuth-authenticated accounts a site namespace derived from the email address, for example [email protected] might correspond to john-smith.chatgpt.site; these namespaces are recorded in public Certificate Transparency (CT) logs along with TLS certificates, so anyone can look them up even if the Site is set to private. OpenAI fixed the vulnerability in the week of October 12, 2026, but CT logs by design retain historical certificate records, and certificates issued before the fix can still be found even if the corresponding Site has been deleted. The issue does not directly expose full email addresses, credentials or private Site content, but the identifiers can be used to construct candidate email addresses, link online identities or assist targeted phishing.