OpenAI announced a brand new function that it says will present extra safety from immediate injection assaults, the place malicious chatbot directions are hidden in webpages and different content material sources.
Amongst different issues, Lockdown Mode will disable reside net shopping (so you possibly can solely entry cached content material), the retrieval and show of photographs from the net (you possibly can nonetheless generate photographs), deep analysis, and agent mode.
The corporate says that even with Lockdown Mode turned on, ChatGPT might nonetheless be weak to immediate injections — which might, for instance, “seem in cached net content material or in an uploaded file, and will nonetheless have an effect on the conduct or accuracy of a response.”
However the purpose is to scale back the chance that delicate knowledge will get shared within the course of.
“Lockdown Mode is not supposed for everybody,” OpenAI says. “It’s designed for individuals and organizations that deal with delicate knowledge and need stricter safety from knowledge exfiltration dangers associated to immediate injection.”
The corporate says it’s at the moment rolling Lockdown Mode out to self-serve ChatGPT Enterprise accounts, in addition to eligible private accounts.
