OpenAI is gradually expanding its security options for ChatGPT users by making Lockdown Mode available to a much wider audience. Initially introduced in February to a limited group of high-profile users such as executives and security teams, the feature now reaches personal accounts across Free, Go, Plus, and Pro tiers, as well as self-serve Business plans. Users can activate it through the settings menu under Security, once it appears for their account.
In practice, Lockdown Mode trades convenience for caution. It restricts or disables several outward-facing capabilities, including live web browsing, which is limited to cached content that may prove outdated or incomplete. Image handling in responses is curtailed, though uploads and generation remain possible where supported. Features like Deep Research, Agent Mode, and Canvas networking are turned off entirely, as is the ability to download files for analysis. The goal is to reduce potential pathways for sensitive data to leave the conversation, particularly in the face of prompt injection attacks. Even so, the mode does not eliminate risks entirely—malicious instructions could still slip through uploaded files or cached material.
This measured approach reflects broader industry realities. As large language models have become everyday tools, concerns around data exfiltration, adversarial inputs, and unintended leaks have grown more pressing. OpenAI’s rollout acknowledges that not every user needs maximum openness, especially those handling sensitive information or operating in regulated environments. Yet it also highlights the inherent tension in these systems: the same connectivity that makes ChatGPT useful for research, coding, and creative work also creates vulnerabilities that are difficult to patch completely.
Complementing the security update, OpenAI has introduced Active Sessions across most account types. This allows users to review active logins, including device details, approximate location, sign-in times, and whether the session is current or trusted. It offers a straightforward way to monitor and revoke access, addressing a common pain point in cloud services where accounts can linger on forgotten devices. Notably, the feature is unavailable for organizations using single sign-on setups like SAML or OIDC, which may limit its utility for larger enterprises.
Taken together, these changes represent incremental progress rather than a transformative leap. OpenAI continues to iterate on safety measures amid ongoing scrutiny of how AI tools handle privacy and security. For many users, Lockdown Mode provides a practical off-ramp when the default experience feels too exposed, while Active Sessions add basic oversight long standard in other productivity platforms. Still, the piecemeal nature of these enhancements underscores how quickly AI capabilities have outpaced robust, universal safeguards.
In a landscape where ChatGPT has moved from novelty to infrastructure for millions, such features matter more than ever. They signal a maturing platform willing to let users dial down functionality for peace of mind, even if it means accepting certain limitations. Whether this balance satisfies growing demands for transparency and control remains an open question as adoption deepens across personal and professional contexts.
