OpenAI is rolling out broader access to two ChatGPT security controls: Lockdown Mode and Active Sessions. Lockdown Mode is designed to mitigate the final stage of data exfiltration in prompt injection attacks by restricting outbound network requests, disabling features like live web browsing and file downloads. It does not block prompt injections themselves. Active Sessions enables users to review and manage devices and sessions where their ChatGPT account is logged in, allowing them to log out of unrecognized sessions. These features are intended to enhance account security, particularly for users with highly sensitive data. OpenAI also offers an Advanced Account Security option that disables password logins in favor of physical security keys or passkeys and improves account recovery mechanisms. These measures collectively strengthen protection against account takeover and data leakage risks.

The features reduce the risk of data exfiltration via prompt injection attacks and unauthorized account access. Lockdown Mode limits capabilities that could be exploited to transfer sensitive data out of the ChatGPT environment. Active Sessions helps users detect and terminate unauthorized sessions, reducing the risk of account compromise. The Advanced Account Security feature further mitigates account takeover risks by removing password-based authentication and enhancing recovery security. There are no known exploits in the wild related to these features at this time.

These security controls are currently being rolled out and can be enabled by users in their ChatGPT account settings. Lockdown Mode should be used by users and organizations handling highly sensitive data who require additional protection against data exfiltration. Active Sessions is available to all ChatGPT users except those using organizational SSO and should be used to monitor and manage active logins. Advanced Account Security is an opt-in feature recommended for users at increased risk of targeted attacks. Since these are newly introduced security features, users should enable them as appropriate to strengthen account security. Patch status is not applicable as these are new security controls rather than vulnerability fixes.