OpenAI launches Advanced Account Security for ChatGPT with mandatory passkeys and disabled AI training

OpenAI launches Advanced Account Security for ChatGPT with mandatory passkeys and disabled AI training

OpenAI has released Advanced Account Security, an opt-in feature for ChatGPT users that requires stronger authentication methods and automatically disables AI training on user conversations.

The feature, announced May 4, 2026, is available to all individual ChatGPT users (free and paid) and is targeted at security-conscious users including political dissidents, journalists, elected officials, and researchers. OpenAI says it expects to expand the feature to enterprise customers.

Four security settings

Advanced Account Security includes four mandatory settings:

1. Passkey or physical security key requirement: Users must authenticate with a passkey or hardware security key instead of passwords. OpenAI requires two authentication methods to be set up. The company partnered with Yubico to offer a bundle of two YubiKeys (YubiKey C NFC and YubiKey C Nano) for $68, down from the usual $126 price. Once enabled, traditional passwords are no longer accepted.

2. Enhanced account recovery: Email and SMS authentication for account recovery are disabled. Users must rely on backup passkeys, security keys, or recovery keys that ChatGPT automatically generates during setup.

3. Shortened login sessions: Active login sessions are shortened to reduce exposure time and require more frequent re-authentication.

4. Automatic AI training opt-out: The feature automatically disables the use of user conversations for AI training. By default, ChatGPT can use conversations to improve its models, but Advanced Account Security turns this off without requiring manual configuration.

Enrollment and availability

Users can enroll through OpenAI's dedicated enrollment web page. After completing setup, users are immediately required to log in again using their new authentication method.

The feature also extends protection to Codex for users developing and fine-tuning code.

According to OpenAI: "As AI becomes increasingly embedded in our lives, it is more important than ever to ensure that users have the controls they need to help protect their privacy and security."

What this means

This represents OpenAI's most significant account security upgrade for individual users, addressing growing concerns about AI systems accessing sensitive data. The mandatory passkey requirement eliminates password-based attacks entirely, while automatic AI training opt-out provides stronger privacy guarantees than the current manual toggle. The Yubico partnership and discounted hardware keys lower the barrier to adoption for security-conscious users. However, the opt-in nature means most ChatGPT users will likely continue with standard password authentication unless OpenAI eventually makes these protections mandatory.