New ChatGPT Lockdown Mode to Mitigate Prompt Injection and Data Exfiltration Attacks

ChatGPT Lockdown Mode is a new security feature designed to mitigate data exfiltration risks by restricting outbound network access during prompt injection attacks. While it disables capabilities like live web browsing and deep research, it does not prevent malicious payloads from entering the model’s context.