OpenAI has begun rolling out Lockdown Mode, an additional security setting designed to provide enhanced user protection against prompt injection attacks.
As artificial intelligence has become better at gathering information from the internet, attackers have begun hiding malicious instructions on web pages and elsewhere to try to outsmart these systems.
OpenAI has begun rolling out Lockdown Mode, an additional security setting designed to provide enhanced user protection against prompt injection attacks.
As artificial intelligence has become better at gathering information from the internet, attackers have begun hiding malicious instructions on web pages and elsewhere to try to outsmart these systems.
OpenAI positions Lockdown Mode as a kind of last line of defense against prompt injections, complementing the robust protections that the company says are already in place in ChatGPT, its models, and internal systems, Engadget writes .
“Block mode is not for everyone,” OpenAI explains. “It is designed for people and organizations that work with sensitive data and need stronger protection against the information leakage risks associated with prompt injections.”
To do this, enabling Lockdown Mode limits some of the features that OpenAI offers in ChatGPT and its other products. For example, you can generate images and upload photos to ChatGPT, but the chatbot will not be able to pull images from the internet or display any images in its response.
Additionally, the AI won't be able to upload third-party files for analysis, though you can still manually attach documents if you want to hear its analytics. Other tools, such as deep research and agent lore, are disabled entirely.
“Lockdown mode does not affect memory, file uploads, the ability to share conversations, or whether your conversations are used to train models,” OpenAI adds. “Many of these settings can be configured separately by workspace administrators.”
The company also notes that the blocking mode will not prevent prompt injections from appearing in content processed by ChatGPT. Instead, it is designed to prevent an attacker from stealing sensitive data from your account by limiting the network requests that someone could exploit.
Lockdown Mode is available for all personal accounts, including those using the free version of ChatGPT from OpenAI.
To activate it, you need to:
You can temporarily disable additional protection, to do this you need:
Separately, OpenAI is implementing an active session manager that allows users to see all the devices or browsers that have been used to access their account. From there, the company offers the option to log out of individual sessions or all at once. Just keep in mind that the latter procedure can take up to 30 minutes to complete.
“If you suspect unauthorized activity on your account, change your password (if you use one), check your login methods, and contact OpenAI support,” the company added.
