OpenAI unveils account security mode

OpenAI has launched Advanced Account Security to reduce the risk of account takeover for high-risk accounts.

• Once enabled, regular password login cannot be used, and users must register 2 physical security keys or a passkey.
• Email/SMS recovery is removed, and recovery is only possible via a recovery key, backup passkey, or physical security key.
• When account protection is turned on, even OpenAI support staff can no longer perform account recovery on the user's behalf, blocking social engineering attacks targeting the support portal.
• Login sessions are kept shorter, notifications are provided upon login, and users can check active sessions on ChatGPT and Codex.
• The model training exclusion option is applied by default.
• OpenAI is partnering with Yubico to offer security key bundles at a low price.
• Participants in the Trusted Access for Cyber program must either turn on this mode by June 1 or prove they are using phishing-resistant authentication based on enterprise SSO.

The target audience includes journalists, elected officials, political dissidents, researchers, and others with security-sensitive accounts.