ChatGPT Gets 'Locked Chats' to Protect Sensitive Conversations

PCMag

Source published: Site published:

Original article · فارسی

OpenAI is adding a new privacy feature to ChatGPT that lets users hide and protect sensitive conversations. This can help keep chats private if someone else uses their phone or computer or can see their screen.

“Many users use ChatGPT in crowded spaces, like a bus or a subway platform, and may not want others around them to be able to see their conversations,” the company explained. “Similarly, it's common to hand your device to friends or family when you want to show them something in a chat.”

To keep sensitive conversations hidden, OpenAI has created a feature called “locked chats.” It works much like WhatsApp’s Chat Lock. The feature removes selected ChatGPT conversations from the regular chat history, pinned chats, and conversation search on the app's main page.

Instead, sensitive chats appear under Projects > Locked chats in the ChatGPT sidebar. The titles of locked chats stay hidden, and each locked conversation is protected by a four-digit account PIN chosen by the user. On mobile devices, users can also protect locked chats with a face scan or fingerprint scan.

(Credit: OpenAI)

To lock or unlock a chat, tap the three-dot menu in a ChatGPT conversation. The feature is being introduced today for ChatGPT Free, Go, Plus, Pro, Business, Enterprise, and Edu accounts, but it has not appeared for everyone yet. OpenAI explains on its support page: “Locked chats are rolling out gradually. If you don’t see Lock chat in an eligible chat’s more options menu (•••), the feature may not be available for your account yet.”

OpenAI also points out that it can still access locked chats and may use them to train its AI models if a customer has chosen to take part through the “Improve the model for everyone” setting. The chats do not use end-to-end encryption. The support page also explains that locking a chat does not “remove information that ChatGPT remembered before you locked the chat” or “override applicable workspace policies, safety requirements, or legal and compliance obligations.”

There is another important restriction. If you forget your PIN and choose to reset it, OpenAI must permanently delete all your locked chats, along with any files and other content protected by the locked chat feature. However, if you know your current PIN, you can change it without losing your locked chats.

Another protection relates to ChatGPT's memory feature. It keeps the information from regular chats separate from information in locked chats. OpenAI explains: “With Memory enabled, your locked chats project has its own memory, shared between the chats inside it. New messages you send after locking a chat can update that memory, but do not update your regular Memory, so that new information from your locked chats doesn’t leak into regular chats.”

The support page lists other limits, including that temporary chats cannot be moved into locked chats.

(Credit: OpenAI)

Disclosure: Ziff Davis, the company that owns PCMag, filed a lawsuit against OpenAI in April 2025, claiming that OpenAI infringed Ziff Davis copyrights when training and running its AI systems.

About Our Expert

Michael Kan
Principal Reporter

Experience

I've worked as a journalist for more than 15 years. I began my career reporting on schools and cities in Kansas City. I joined PCMag in 2017 and cover satellite internet, cybersecurity, computer hardware, and other topics. I now live in San Francisco. Before that, I spent more than five years in China reporting on its technology industry.

Since 2020, I've covered the launch and rapid growth of SpaceX's Starlink satellite internet service. I've written more than 600 stories about where the service is available and new features. I've also reported on legal and regulatory fights over the growth of satellite constellations, competition with providers such as AST SpaceMobile and Amazon, and efforts to bring satellite service to mobile phones. I've studied Federal Communications Commission (FCC) records to find news and traveled to remote parts of California to test Starlink's cellular service.

I also report on online security threats, including ransomware groups and the rise of malware powered by AI. In 2024 and 2025, the Federal Trade Commission (FTC) forced Avast to pay consumers $16.5 million after the company secretly collected and sold their personal information to other companies. I reported on this in a joint investigation with Motherboard.

I also cover the computer graphics card market. During shortages caused by the pandemic, I camped outside a Best Buy to get an RTX 3000 graphics card. Now I'm following how shortages of memory chips, driven by demand for AI, are affecting the entire consumer electronics market. I'm always eager to learn more, so please share feedback in the comments and send me tips.

Read Full Bio

Other sources for this story