OpenAI has introduced Advanced Account Security, a new opt-in setting designed to safeguard ChatGPT and Codex accounts for individuals at increased risk of digital attacks. The security suite brings together phishing-resistant sign-in methods by requiring passkeys or physical security keys while disabling traditional password-based logins. To reduce risks associated with compromised emails or phone numbers, the feature also disables email and SMS recovery in favor of backup passkeys and recovery keys, meaning OpenAI Support cannot assist with account recovery for enrolled users. Additionally, the setting shortens sign-in sessions, provides login alerts, and automatically excludes conversations from model training. To help users adopt hardware-based authentication, OpenAI partnered with Yubico to offer preferred pricing on customized security key bundles, though any FIDO-compliant security key or software-based passkey can be used. Furthermore, individual members of Trusted Access for Cyber accessing advanced models will be required to enable Advanced Account Security starting June 1, 2026, unless their organization attests to using phishing-resistant authentication via single sign-on. OpenAI plans to extend these protections to additional audiences, including enterprise environments, over time.
Introducing Advanced Account Security
OpenAI has introduced Advanced Account Security, a new opt-in setting for ChatGPT and Codex accounts designed to protect high-risk users and those seeking enhanced safeguards. The feature combines phishing-resistant sign-in, stricter recovery methods, and automatic training exclusion for sensitive conversations. Additionally, OpenAI has partnered with Yubico to offer discounted hardware security keys.
Why it matters
As AI tools increasingly handle sensitive professional and personal data, robust account security measures are critical to prevent unauthorized access and data breaches for high-risk users.
Topic dossiers OpenAI