ChatGPT Privacy Center and Security History Explained: Two New Ways to Check Your Account

OpenAI shipped two account-checking surfaces four days apart in late September 2026, and neither got much attention because neither is an AI feature. Privacy Center, on September 21, puts everything ChatGPT does with your data in one place, and it is on the free plan. Security history, on September 25, answers a question a lot of people have quietly wondered about: has anybody else signed into my ChatGPT? Here is where each one lives and what it actually tells you.

Key Takeaways

• Privacy Center is free. OpenAI: it is "rolling out to signed-in ChatGPT Free, Go, Plus, and Pro users."
• It is a map, not a new control panel. OpenAI: "Your privacy controls remain in ChatGPT Settings." Privacy Center explains and links; it does not replace the switches.
• It covers six areas: "chat privacy, memory, personalization, data use, connected apps, and account security."
• It is hidden under Help on the web, not under Settings. Account menu, then Help, then Privacy Center.
• Security history lists sign-ins and sign-outs, plus "changes to multi-factor authentication (MFA), passkeys, and other security settings."
• Each event carries time, location and device details -- which is what makes an intrusion visible rather than theoretical.
• OpenAI hedges the detail quality itself: "Some details may be approximate or unavailable." A location that looks wrong is not proof of a break-in.
• OpenAI named no plan for security history. Unlike Privacy Center, the release note carries no plan list, so we are not asserting one.

Where to Find Each One

Two different menus, which is the main reason people will not find them by accident. These are OpenAI's own paths, from its release notes.

Two honest gaps in that table. OpenAI describes the security history path for "ChatGPT on the web" and does not name an iOS or Android route, so on mobile treat it as unconfirmed rather than absent. And Privacy Center is the rare ChatGPT feature that lives under Help on the web rather than Settings, which is worth remembering because it is the last place most people look for a privacy page.

Privacy Center: What It Is For

It is the answer to "what does ChatGPT actually do with what I type?", assembled in one place instead of scattered across six settings screens. OpenAI's description: Privacy Center "brings together information about chat privacy, memory, personalization, data use, connected apps, and account security, with links to the settings that manage them."

The important sentence is the next one: "Your privacy controls remain in ChatGPT Settings." Nothing moved. If you have turned off model training, or set memory to off, or disconnected an app, those switches are where they were. Privacy Center is documentation with shortcuts attached.

That is less exciting than a new control, and more useful than it sounds. The six areas it covers are exactly the ones this site gets asked about, and each of them has its own separate behaviour that is easy to get wrong:

• Chat privacy and what a temporary chat does and does not keep.
• Memory, including the project-scoped kind.
• Personalization, which is a different setting from memory.
• Data use, meaning whether your conversations train future models.
• Connected apps, which is where third parties get read access.
• Account security, which is now where security history sits.

OpenAI adds the usual caveat: "Available options depend on your plan, region, and workspace settings." On a work or school account an administrator's settings sit on top of yours, and in some regions the defaults differ.

Security History: Checking Whether Someone Else Has Been In

This is the more practically useful of the two, and the shorter to explain. OpenAI: "We're introducing security history, a new way to review recent security activity for your OpenAI account."

What gets logged, in OpenAI's words:

• Sign-ins and sign-outs
• Changes to multi-factor authentication (MFA)
• Changes to passkeys
• Changes to other security settings

And for each one: "Events include the time, location, and device details."

That combination is what makes the feature worth five minutes of your attention. A ChatGPT account often holds years of conversations, uploaded documents, connected email and calendar, and on some plans connected bank accounts. Until September 25 there was no consumer-facing way to check whether a session you did not start existed. Now there is a list.

One caution OpenAI writes itself, and it matters if you are the anxious type: "Some details may be approximate or unavailable." Location in a log like this is usually derived from an IP address, which can place you in the wrong city, show a mobile carrier's hub rather than your street, or move because you used a VPN. An unfamiliar city on its own is weak evidence. An unfamiliar device, or a sign-in at a time you were asleep, is much stronger.

What to Do If You See Something You Do Not Recognise

Treat it as a compromised password until proven otherwise, and work in this order. OpenAI points to its "Keeping your OpenAI account secure" help article for this, and the two protections that show up in security history are the two worth turning on:

1. Change your password. This is the only step that ends an attacker's existing access on its own.
2. Turn on multi-factor authentication, so a stolen password is no longer enough by itself.
3. Consider a passkey, which replaces the password with your device's own unlock.
4. Review connected apps, because an app you authorized keeps its access regardless of your password.
5. Check security history again in a day, since any of the above that you did will now appear in it as a dated event, and anything you did not do will stand out against them.

The nice property of the design is that step 5 works: because the log records MFA and passkey changes too, your own remediation leaves a trail you can recognise.

Should You Bother?

Open security history once, today, if your ChatGPT account has anything in it you would not post publicly. It takes under a minute, it costs nothing, and the only reason not to is that until this week it did not exist.

Open Privacy Center if you have ever wondered whether your chats train the models. That is the single question it is best placed to answer for your specific account, plan and region, rather than in the abstract.

Skip Privacy Center if you have already set your controls deliberately. It adds no switches. If you have turned off training and memory and you know where both live, there is nothing here you do not have.

Do not read either one as a plan upgrade. Privacy Center is explicitly on Free and Go as well as Plus and Pro. Neither surface is a reason to pay, and if you are weighing that decision, whether ChatGPT Plus is worth it is the page for it.

Sources

• OpenAI Help Center, "ChatGPT Release Notes", the entry "Privacy Center in ChatGPT" dated September 21, 2026 and the entry "Security history in ChatGPT" dated September 25, 2026. Every quoted line here comes from those two entries, read September 26, 2026.
• OpenAI links a dedicated "Privacy Center in ChatGPT" help article and a "Keeping your OpenAI account secure" article from those entries. Neither had a readable archived copy on September 26, 2026, so nothing on this page rests on them; we will add their detail once they can be read.