OpenAI rolled out an optional advanced security feature for ChatGPT users this week. The feature replaces passwords with passkey authentication and sharply limits account recovery options. Users who enable it also get automatic exclusion of their chats from OpenAI's training data pipeline.
Passkey Authentication Mechanics
Users must sign in using their device's biometric sensor or PIN instead of a password. This happens locally on the device—no credentials travel to OpenAI's servers. It stops phishing attacks cold. But if you lose your phone and didn't save the recovery code, you're locked out permanently. OpenAI doesn't support email-based account recovery for these accounts.
Turning this on means you're on your own for access. There's no safety net. Save that recovery code in your password manager or a locked drawer. Forget it, and the account vanishes.
Training Data Privacy Shift
Chats in secured accounts won't feed into OpenAI's model training. The company doesn't collect those conversations for future improvements. This happens automatically when you enable the security feature—no extra opt-out step needed.
Standard ChatGPT accounts still share chat data unless users manually toggle off training data sharing in settings. This new setup bundles privacy with security from the start. It's a single toggle for both protections.
Who Should Enable It
High-value accounts like business users or security teams should consider it. The trade-off is clear: stronger protection against account takeovers versus real lockout risk. Most casual users won't need it. Standard security remains adequate for everyday use.
OpenAI hasn't shared adoption numbers. The feature's value depends entirely on whether you can securely store recovery codes. If not, stick with traditional logins. This isn't for the careless.
Users can activate the feature immediately in ChatGPT account settings under security options.
