OpenAI’s Lockdown Mode: what does it do?

OpenAI rolls out “Lockdown Mode” for prompt injection

OpenAI has introduced an optional security setting called Lockdown Mode, designed to offer users advanced protection from prompt injection attacks. The idea is to reduce exposure by limiting some features that can be abused when malicious instructions are injected into prompts or tool-using workflows.

The rollout positions Lockdown Mode as something most users may not need day-to-day, but which can be valuable for higher-risk use cases—especially those that involve untrusted inputs or environments where attackers might attempt to redirect the model’s behavior.

Why prompt injection protection matters

Prompt injection is a class of attacks where adversaries craft text (or related inputs) to manipulate an AI system into doing something unintended—such as revealing sensitive information, following attacker-supplied instructions over system policies, or triggering unsafe tool usage. As models increasingly support richer interactions, including actions and retrieval, the blast radius of prompt injection increases.

Lockdown Mode’s relevance is that it treats security as a user-selectable posture rather than a one-size-fits-all configuration. By gating capabilities, OpenAI is effectively trying to reduce the attack surface when the user opts into stronger protections.

What users should take away

• It’s an optional security setting
• It targets prompt injection threats
• It works by restricting certain features

The feature’s availability is part of OpenAI’s ongoing effort to give users more control over how defensive the system behaves, particularly when handling risky or untrusted content.