OpenAI rolls out Lockdown Mode to curb ChatGPT data‑exfiltration

OpenAI has introduced an optional security setting called Lockdown Mode for ChatGPT. When enabled, the mode disables a range of internet‑connected capabilities – live web browsing, image retrieval from the web, Deep Research, Agent Mode, Canvas‑generated code and file downloads – limiting the model to cached content, uploaded files and built‑in image generation. The feature is intended for users and organisations that handle sensitive information and want to reduce the risk of data exfiltration via prompt‑injection attacks. It is available to free, Go, Plus, Pro, Business and Enterprise accounts, and can be activated in Settings → Security → Advanced Security. OpenAI stresses that Lockdown Mode does not prevent prompt‑injection content from appearing in inputs, but it blocks the final step of sending extracted data to external networks. The rollout is global and is marketed as a “last line of defence” for high‑value use cases such as legal, medical or financial data handling, while acknowledging that it is a trade‑off between functionality and security.