Just OpenAI announced Lock mode. Remember the lockdown? It’s like that, but with an artificial intelligence chatbot.
In fact, it was introduced as a way to protect you from emergency injection attacks –is an insidious new type of AI-centric spammer activity that is sometimes referred to as full-blown hacking.. Attackers may be able to not only steal data, but also exploit it by placing malicious hints in content that is queried or encountered by an AI agent. to control the user’s life.
Lockdown mode is, as the name suggests, a toned down experience. It’s not automatically a threat-induced lockout, but a kind of panic room you enter when using ChatGPT if what you’re doing is so sensitive that an emergency injection attack would be particularly damaging.
In fact, any time LLM breaches the security of a small chatbot window in your browser or smartphone app and browses the web for information or retrieves an image or – probably – tries to buy an airline ticket on your behalf, it puts you at risk of an emergency injection attack, so it blocks those features.
Lockdown mode means that ChatGPT cannot:
- browse the web
- show images in replies (but it can generate images and you can upload images)
- Do “Deep Research”.
- acts as an agent
- Network with Canvas code generator
- download the files
As OpenAI says:
“It’s lock mode no it is meant for everyone. It is designed for people and organizations that manage sensitive data and want to protect against data exfiltration risks associated with operational injection.
It’s early days with AI chatbots, and ChatGPT is creating a regime to prevent data theft. That makes sense, I think. Still, it’s frustrating when, for example, lawyers dump sensitive client data into ChatGPT, or doctors upload their patients’ health data into their favorite LLM. To protect sensitive data from emergency injection attacks, there is a safer alternative “lockdown mode”: do not allow sensitive data near the chatbot.
