OpenAI Introduces Lockdown Mode to Mitigate Prompt Injection Risks OpenAI has announced the release of Lockdown Mode, a new security feature designed to protect users from prompt injection attacks. These vulnerabilities occur when malicious instructions are concealed within webpages, documents, or other external content sources to manipulate AI behavior. The update restricts several dynamic capabilities to minimize data exfiltration risks. When activated, the mode disables live web browsing, restricting the model to cached content only. It also blocks the retrieval and display of images from the internet, while offline image generation remains available. Features such as Deep Research and Agent mode are simultaneously deactivated. OpenAI explicitly notes that Lockdown Mode does not provide absolute immunity. The company acknowledges that prompt injections may still appear in cached web content or uploaded files, potentially influencing response accuracy. The primary objective is to significantly reduce the probability of sensitive information being shared or exfiltrated during interactions. The feature is targeted specifically at enterprise users and individuals managing confidential data. OpenAI states that the mode is not intended for general consumers, but rather for organizations requiring stricter safeguards against data leakage. The rollout is currently underway for self-serve ChatGPT Business accounts and eligible personal subscriptions. By limiting external data retrieval and autonomous browsing capabilities, the update represents a measured approach to balancing AI functionality with enterprise-grade security. Organizations handling proprietary or regulated information can now deploy a controlled environment that reduces exposure to third-party content manipulation. Further enhancements to prompt injection defenses are expected as OpenAI continues to refine its security infrastructure.