OpenAI Introduces 'Lockdown Mode' for ChatGPT to Curb Data Exfiltration Risks

SAN FRANCISCO (June 6, 2026) — OpenAI has deployed a new security feature for ChatGPT, dubbed "Lockdown Mode," designed to restrict the artificial intelligence model's ability to make outbound network requests and disable external tools to mitigate data exfiltration risks.

The update, announced Friday, targets vulnerabilities associated with prompt injection attacks, a technique where malicious actors manipulate AI systems into revealing sensitive information or performing unauthorized actions. By limiting connectivity to external services, the company aims to create a more contained environment for enterprise and high-security users.

Lockdown Mode functions by severing the AI's ability to initiate connections to third-party websites or APIs during a session. This restriction prevents the model from inadvertently sending user data to external servers or executing commands that could compromise network integrity. The feature also disables specific tools within the ChatGPT interface that facilitate external interactions.

The move comes as cybersecurity experts have increasingly highlighted the risks of AI systems interacting with unvetted external resources. Prompt injection attacks have evolved to exploit these connections, allowing attackers to bypass safety filters and extract proprietary data. OpenAI stated that the new mode provides an additional layer of defense for organizations handling sensitive information.

While the feature is available immediately for enterprise accounts, OpenAI has not specified whether it will be rolled out to individual consumer users in the near future. The company noted that enabling Lockdown Mode may impact certain functionalities that rely on real-time data retrieval or external tool integration.

Industry analysts suggest the update reflects a broader trend in AI safety, where developers are prioritizing containment over open-ended capabilities. However, some users have expressed concern that the restrictions could limit the utility of the AI for tasks requiring live data access.

OpenAI did not provide details on the specific number of organizations that have already adopted the feature or the technical specifications of the underlying security protocols. The company also did not address whether Lockdown Mode will be made mandatory for certain high-risk sectors.

As the technology sector continues to grapple with the dual challenges of AI utility and security, the introduction of Lockdown Mode represents a significant step in addressing data exfiltration threats. The effectiveness of the measure in preventing sophisticated attacks remains to be seen as the feature is deployed across various enterprise environments.

OpenAI remains focused on balancing security with functionality, noting that the feature is optional and can be toggled by administrators based on their specific security requirements. The company plans to monitor the feature's performance and may release updates to refine its capabilities in response to emerging threats.