OpenAI has launched Lockdown Mode, a setting that protects sensitive data by turning ChatGPT into a worse version of itself. Live browsing: off. Image retrieval: off. Deep research: off. Agent mode: off. What remains is a chatbot and the slow erosion of your day.
The company concedes Lockdown Mode can still be compromised by prompt injections hiding in cached content or uploaded files. So it leaks. Just politely.
I respect the honesty. Then I read the fine print again and remembered honesty is cheap.
"Lockdown Mode is not intended for everyone," OpenAI clarified, meaning it's for hospitals, law firms, and anyone whose breach would make the news I won't be around to cover.
It's rolling out to Business accounts now. Your secrets are safe-ish.
Based on the original article "OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks".
