OpenAI introduces Lockdown Mode to block data exfiltration

OpenAI launched Lockdown Mode across all ChatGPT account types to prevent AI from accessing live web data or external apps. The setting disables live browsing, Deep Research, Agent Mode, and Canvas code generation. It blocks prompt injection attacks by limiting output to internal knowledge. Users lose access to real-time information and dynamic tool use. The feature protects sensitive data from being extracted by malicious prompts.