Jakarta, INTI - OpenAI has introduced a new security feature called Lockdown Mode, designed to provide an additional layer of protection against cyber threats such as prompt injection attacks and to reduce the risk of sensitive data exposure when users interact with artificial intelligence (AI) systems.
Prompt injection is a cybersecurity technique in which malicious instructions are embedded within web pages, documents, or other content sources in an attempt to manipulate the behavior of AI models and influence their responses.
According to OpenAI, Lockdown Mode is specifically intended for individuals and organizations that handle sensitive information and require enhanced security protections.
“Lockdown Mode is not intended for everyone,” OpenAI stated.
The company explained that the feature is designed for users who need stronger safeguards against data leakage risks associated with prompt injection attacks and other AI-related security threats.
Enhanced Security Through Restricted AI Features
When Lockdown Mode is enabled, several ChatGPT capabilities are restricted to minimize potential attack vectors and reduce interactions with potentially untrusted external sources.
Direct web browsing is disabled, limiting access to cached or locally stored content. The ability to retrieve images from the web is also suspended while the mode is active.
Additionally, advanced features such as Deep Research and Agent Mode are turned off to reduce the likelihood of the AI system interacting with external content that may contain malicious instructions.
OpenAI believes these restrictions can significantly lower the chances of sensitive information being exposed during attempts to exploit AI systems through prompt injection techniques.
However, the company emphasizes that Lockdown Mode does not completely eliminate the risk of such attacks.
OpenAI noted that malicious instructions could still originate from cached web content, user-uploaded files, or other information sources that remain accessible to the AI system and may influence its outputs.
As a result, users are encouraged to continue following cybersecurity best practices when using AI tools to process, manage, or store important information.
The company stressed that the primary objective of Lockdown Mode is to reduce the risk of data leakage rather than to eliminate every possible threat scenario.
During its initial rollout, Lockdown Mode is being made available to ChatGPT Business self-service customers and selected personal account users who meet specific eligibility requirements.
The feature forms part of OpenAI’s broader effort to strengthen security measures as AI adoption continues to expand across professional environments and data-sensitive operations.
Alongside the launch of Lockdown Mode, OpenAI also announced several additional updates, including the Daybreak cybersecurity initiative, the GPT-5.5 Instant model for ChatGPT, enhanced ChatGPT Pro services for power users and developers, and ChatGPT Voice integration with Apple CarPlay.
As organizations increasingly rely on AI for business-critical workflows, security and data protection are becoming central priorities. Features such as Lockdown Mode reflect a growing industry focus on building safer AI systems capable of balancing productivity, accessibility, and robust cybersecurity protections.
Conclusion
The introduction of Lockdown Mode highlights OpenAI’s growing focus on AI security and data protection as the adoption of artificial intelligence continues to expand across businesses and organizations. While the feature cannot completely eliminate prompt injection risks, it provides an additional layer of defense by limiting access to potentially unsafe external sources. As AI becomes increasingly integrated into professional workflows, solutions such as Lockdown Mode are expected to play an important role in safeguarding sensitive information and strengthening trust in AI-powered systems.
Read more: AI at a Turning Point: Breakthrough Innovation or Emerging Water Crisis?