ChatGPT has been locked down: OpenAI strengthens user protection

The OpenAI company announced the official introduction of a special blocking mode under the code name “Lockdown Mode” into the functionality of ChatGPT and other related technological products. This technological innovation acts as an auxiliary protective tool, aimed at minimizing the risks of so-called prompt injection (implementation of prompts) — a class of security threats focused mainly on large speech AI systems.

 Lockdown Mode in ChatGPT

The Lockdown Mode security system is a functional limiter extension created as the ultimate line of defense for AI infrastructure against manipulative-class attacks (implementation of third-party directives). Hackers practice prompt injections in order to force a model to remotely share system correspondence details or confidential information located on a protected terminal or in a user’s chat memory.

The new Lockdown Mode works to pre-empt the threat of ultimate session hijacking by purposefully restricting outgoing background web requests from chatbots to the outside environment, depriving a potential malicious agent of the tools to exfiltrate a user’s personal data.

Technical limitations

When Lockdown Mode is enabled, the AI cuts out some of the useful interactive interaction an attacker needs to work out network bypasses. Features of operation when switching the safety slider include:

• Disabling offline internet artwork: ChatGPT allows you to seamlessly work with manually uploaded local footage and create quality art from scratch, but removes the technical ability to download content from addresses on the World Wide Web or host it as media in an internal chat block.
• External Analysis Isolation: The security system automatically prohibits external access to the network for any automatic text/archive downloads, while no one prevents users from manually downloading documents or files locally.
• Deep Service Freeze: Additional specialized capabilities, including Deep Research and Agent Mode, cease their own existence until the Lockdown Mode secure area is forced to close.

It is also worth fixing the key parameters: the current introduction of the protective barrier does not in any way interfere with the built-in elements of the virtual mentor’s standard memory, does not reduce the settings for free download of files and distribution (sharing) of log correspondence. The administrative branch of large group business licenses always retains absolute freedom to choose options for its network in the interface settings.

Active Session Manager – Cyber Security Support Update

In addition to the launch of Lockdown Mode, OpenAI engineers further optimized the IT security functional stack of users by implementing a built-in interactive manager of active Internet sessions. It displays comprehensive current activity: input device hardware system type, exact installed browser system explorer.

The activity manager allows the administrator to log out from his local smartphone at once from the entire ecosystem of web authentications at once or to point-disconnect a “foreign” account in a single list. Closing the desktop web session detection logs on the company’s server platform takes approximately 30 minutes. If suspicious phishing activity is detected, technical support engineers call to reset the user’s master password.

Connect function

Hidden Cyber ​​Fraud Control is added free of charge by the creators to any personal user account (not even requiring a Plus subscription). The scheme of starting the tool is extremely transparent:

• Log in to your own ChatGPT desktop.
• Click on the tab of personal options “Settings”.
• Go through the system window to the “Security” submenu adjustments branch.
• Check the box or click the Lockdown Mode trigger link available inside the Advanced Security section.