ChatGPT Introduces Lockdown Mode to Everyone, Preventing Prompt Injection Attacks

OpenAI has extended its powerful privacy feature, Lockdown Mode, to all ChatGPT users, including those on the free tier. The feature, which helps to avoid malicious actors seizing the AI with the help of the so-called prompt injection attacks, is now offered to Free, Plus, Pro, Go, and self-serve Business accounts.

Lockdown Mode was initially launched in February to select enterprise users. When turned on, it does not allow ChatGPT to issue live outbound network requests. This prevents hackers who may be injecting webpages and content with covert cues that seek to deceive the AI into transmitting personal information to third parties without the user being aware of it.

While Lockdown Mode enhances privacy and security, it does come with trade-offs. The feature blocks or restricts certain ChatGPT features that rely on external calls. The trade-off is usually worth the extra security to privacy-conscious users who handle sensitive data.

OpenAI has updated the original blog post to announce the wider rollout, saying: “Lockdown Mode is now rolling out to personal ChatGPT accounts and self-serve ChatGPT Business accounts. Lockdown Mode was first introduced on ChatGPT enterprise plans, but is now an optional setting that people and teams can use when working with sensitive information or related features.

The relocation is timed by the fact that prompt injection attacks are becoming an increasing issue in the AI sector. In contrast to the conventional hacking techniques, which exploit software vulnerabilities, prompt injection exploits the instructions of the AI by placing malicious commands in text, which the AI reads. In the absence of features such as Lockdown Mode, a user might inadvertently cause an attack by asking ChatGPT to summarize a malicious webpage or document.

Lockdown Mode has now become an optional feature that allows users to sacrifice some functionality to have more control over their data and privacy.