Researchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User Interaction

Cybersecurity researchers have found a vulnerability in OpenAI ChatGPT's Lockdown Mode that could be exploited to bypass its protections. Lockdown Mode is a setting introduced to reduce the risk of data exfiltration from prompt injections. The vulnerability allowed for the potential theft of chat data without user interaction. This flaw could also enable phishing attacks. The issue has now been patched by OpenAI.