OpenAI introduces Lockdown Mode, a new protective feature against attacks based on artificial intelligence.

The producer of ChatGPT has announced a new security feature called Lockdown Mode, designed to protect users against 'prompt injection' attacks. These attacks involve malicious instructions hidden in commands that can compromise sensitive information.

Lockdown Mode disables features that could be exploited in such attacks, such as access to images from the internet, downloading files, and options like Deep Research and Agent Mode. This feature is specifically intended for individuals and organizations that handle confidential data and want stricter protection, thus accepting the limitation of access to certain facilities. The company emphasizes that Lockdown Mode will not prevent the display of injected content but will protect users from the extraction of sensitive information.

The feature will be available to all users, including those without a subscription, and can be activated from the Safety and security section of the ChatGPT application.