MorseSign inStart freeOpen Morse

Is ChatGPT encrypted? In transit and at rest, but not end to end

The Morse teamPublished 8 min read

Yes, ChatGPT is encrypted as of November 2026, but not end to end. OpenAI encrypts your chats in transit with TLS 1.2 or higher and at rest with AES-256. OpenAI holds the keys, and only Enterprise and API customers who set up Enterprise Key Management add a second layer of keys they hold themselves. OpenAI's servers decrypt each message so the model can answer it, so the encryption keeps outsiders out but not OpenAI.

We read OpenAI's trust portal, its enterprise privacy and security pages, its four help articles on Enterprise Key Management, its Locked Chats, shared links, apps, consumer data, and data residency help pages, its US privacy policy, and its November 2025 post on the New York Times lawsuit, all on November 18, 2026. OpenAI deserves credit for how much it puts in writing. Its key management FAQ says that cutting off your own key leaves the data "shredded," with no way back.

ChatGPT encryption: the key facts

LayerWhat OpenAI says, as of November 18, 2026PlansSource
In transitTLS 1.2 or higher "on all endpoints," between you and OpenAI and between OpenAI and its service providersAllOpenAI trust portal
At rest"All customer data is encrypted at-rest using AES-256"AllOpenAI trust portal
End to endNo plan offers it. Locked Chats does not "create end-to-end encryption"NoneOpenAI
Who holds the keysOpenAI, for "our standard data encryption," even when EKM adds a second layerAllOpenAI trust portal
Your own keysEnterprise Key Management adds a second layer with keys in your own key serviceEnterprise agreement, and the APIOpenAI
Client-side encryptionOn OpenAI's "long-term roadmap" since November 12, 2025Not offeredOpenAI

Is ChatGPT end-to-end encrypted?

No, on any plan. End-to-end encryption means only the two ends of a conversation hold the keys, and in ChatGPT the other end is OpenAI's own servers, where the model reads your message in plain text. OpenAI's security page describes encryption "at rest and in transit between you and OpenAI, and between OpenAI and its service providers" (OpenAI).

OpenAI has said it wants to go further. On November 12, 2025, its chief information security officer, Dane Stuckey, wrote that the company's "long-term roadmap includes advanced security features designed to keep your data private, including client-side encryption for your messages with ChatGPT" (OpenAI). He said those features would keep conversations "inaccessible to anyone else, even OpenAI," and were "in active development." A year later, none of the OpenAI pages we read describes client-side encryption in ChatGPT.

The closest thing OpenAI has shipped is Locked Chats, and its help article says locking "does not create end-to-end encryption or a separate encrypted vault" (OpenAI). We go through what the PIN hides in are ChatGPT Locked Chats encrypted.

How does ChatGPT encrypt your chats?

In two places, the wire and the disk. On the wire, OpenAI says all customer data is encrypted "using TLS 1.2 or higher on all endpoints" (OpenAI trust portal). The same rule covers the hop from OpenAI to the companies it uses to run ChatGPT. On disk, OpenAI says "All customer data is encrypted at-rest using AES-256" (OpenAI trust portal). Its enterprise privacy page repeats both standards and adds that OpenAI "uses strict access controls to limit who can access data" (OpenAI).

OpenAI's security page lists the same encryption for individuals and for businesses (OpenAI). OpenAI's US privacy policy, updated September 10, 2026, still warns that "no Internet or email transmission is ever fully secure or error free," and tells you to "take special care in deciding what information you provide" (OpenAI).

Who holds the keys to your ChatGPT chats?

OpenAI does, unless your company has an Enterprise agreement or API account and has set up Enterprise Key Management (EKM). The trust portal describes EKM as encrypting content "twice," once with keys the customer owns and "once with our standard data encryption" (OpenAI trust portal). Without EKM, that standard layer is the only one.

EKM works with a customer's own key service in AWS KMS, Google Cloud, Azure Key Vault, or Oracle Cloud's OCI Vault (OpenAI). OpenAI's help pages describe how it runs:

  • Your master key stays with you. The key encryption key "always stays outside of OpenAI systems" (OpenAI).
  • OpenAI still decrypts. "To read the data, OpenAI takes the eDEK, requests your KMS to decrypt it into a DEK, and then decrypts your data" (OpenAI).
  • Working keys sit in OpenAI's memory. OpenAI caches the data keys "in memory only," and they are "never written to storage" (OpenAI).
  • Revoking is final. Revocation "typically takes effect within one hour," OpenAI says the data is then effectively "shredded," and a workspace can't be recovered by attaching a new key (OpenAI).
  • There's no backout plan. OpenAI says that once a workspace uses customer-managed keys, the only way to stop is a new workspace (OpenAI).
  • Some features drop out. Features in beta or alpha aren't available with EKM on (OpenAI).

So EKM gives a company a kill switch over its stored data, and it doesn't stop OpenAI reading that data while the key is live. After a revocation, OpenAI's key lifecycle page says, Enterprise users "will no longer be able to read previous conversations" either (OpenAI).

Is ChatGPT Enterprise encrypted differently from Plus?

It uses the same TLS and AES-256, with two controls on top. EKM is the first. The second is data residency, which lets eligible API customers and new ChatGPT Enterprise and Edu customers keep content "stored at rest in a specific geographic region," and lets some Enterprise and Edu workspaces keep "model inference (GPU execution)" in that region too (OpenAI).

Neither control limits who in your company can read your chats. On ChatGPT Business, OpenAI says workspace admins "can view, access, export, and delete end user conversations in the workspace" (OpenAI). Our guide to whether ChatGPT Business is private covers what an admin and OpenAI can each see.

What does ChatGPT's encryption protect you from?

From outsiders on the network and from someone who walks off with a disk. It does much less against the people and processes that already have the keys. We set each threat against what OpenAI's own pages say.

ThreatDoes TLS plus AES-256 stop it?What OpenAI says
Someone on your WiFi or your internet providerYes, for the wordsTraffic is encrypted with TLS 1.2 or higher (trust portal)
A stolen drive from a data centerYesStored data is encrypted with AES-256 (trust portal)
OpenAI staff and contractorsNoAuthorized staff and "trusted service providers" may access content for abuse, security incidents, support, legal matters, and model improvement (unless you've opted out) (OpenAI)
A court order or government requestNoOpenAI may share data with government authorities "if required to do so to comply with a legal obligation" (privacy policy)
A bug in ChatGPT itselfNoIn March 2023, a bug "allowed some users to see titles from another active user's chat history" (OpenAI)
Your company's adminsNoBusiness admins can view and export workspace conversations (OpenAI)

In the New York Times dispute, OpenAI itself warned that if the Times won its demand for 20 million conversations, "we will be forced to hand over the very same data we're protecting" (OpenAI). Encryption at rest doesn't change that, because OpenAI can decrypt what it's ordered to produce. How that case played out is in the ChatGPT court order timeline. For staff access, OpenAI's consumer data page says it logs "all access to user content," and asks: "Please do not enter sensitive information that you would not want reviewed or used" (OpenAI).

Are ChatGPT shared links, apps, and mobile apps encrypted?

The connection is, but OpenAI's encryption stops protecting a chat once it leaves your account.

  • Shared links. A link travels over TLS like everything else, but OpenAI says "Anyone who has a link created from a personal account can view the shared conversation" (OpenAI). Links have no expiration date, and "Anyone who can access a link can also forward it."
  • Apps and connectors. Once enabled, an app "may access and use relevant context" from your chats, and may see your IP address and approximate location (OpenAI). From then on, OpenAI says, "Data shared with apps is handled according to each app's terms of service and privacy policies," so OpenAI's encryption promises no longer describe that copy.
  • Mobile and desktop apps. OpenAI's trust portal has a section on the ChatGPT mobile and desktop apps, but its details, listed as "Certificate Pin Exceptions," are shown only to visitors who log in (OpenAI trust portal). ChatGPT Enterprise organizations can also use ChatGPT for Intune, a separate iOS and iPadOS app that lets IT teams apply Microsoft's app protection and Conditional Access policies (OpenAI).

We cover what a router or internet provider can still see in can the WiFi owner see your ChatGPT, and how storage encryption with your own key differs from E2EE in what is zero-access encryption.

How Morse gives you an AI chat that's end-to-end encrypted

ChatGPT's encryption ends at the computer running the model, and OpenAI holds that computer and the keys to everything stored behind it. Staff access, court orders, and bugs all reach a ChatGPT chat because OpenAI can decrypt it. Morse's E2EE mode moves the decryption into a sealed enclave that your own browser checks first, so our servers relay text they can't read.

Pick the mode for what the chat holds:

  • E2EE, for anything you wouldn't type into ChatGPT. Your browser checks the enclave's cryptographic proof itself, then encrypts the message on your device for that enclave alone. Morse's servers relay ciphertext they can't read, and the reply is decrypted only in your browser, with a lock on each reply showing every check that passed. Open-weight models like GLM, Kimi, and DeepSeek answer inside enclaves run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE has no web search or image generation, since models there can use only tools that run on your device.
  • Private, for everyday questions you'd rather no host kept. Open-weight models, plus xAI's Grok, answer on hosts that read the message to reply. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • Anonymous, for GPT, Claude, and Gemini without your account attached. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy, so leave out names and details that point back to you.

In every mode, your chat history is stored in your browser, encrypted, not on Morse's servers. That storage encryption isn't end to end, and our privacy policy says so plainly: "we hold the key, but we never receive the data it unlocks." Morse doesn't train AI on your chats, sell your data, or show ads. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or losing the device deletes your chats, and we can't recover them, so keep a copy of anything you need.

You can start free, and E2EE chats come with Pro, at $19 a month.

Common questions

Is ChatGPT encrypted the way an end-to-end encrypted messenger is?

No. In an end-to-end encrypted messenger, only the devices at each end hold the keys. In ChatGPT, OpenAI's own standard encryption protects stored chats, and OpenAI's servers decrypt each message so the model can answer it.

Can OpenAI employees read my ChatGPT chats if they're encrypted?

Yes, for set reasons. OpenAI says a limited number of authorized staff and service providers may access content to investigate abuse or security incidents, give support, handle legal matters, or improve models unless you've opted out, and that it logs all access to user content.

Does locking a ChatGPT chat encrypt it?

Not beyond the encryption every chat already gets. OpenAI's help center says locking a chat does not 'create end-to-end encryption or a separate encrypted vault.'

Does Enterprise Key Management make ChatGPT end-to-end encrypted?

No. With EKM, OpenAI still decrypts your data by asking your key service to unwrap the key it needs. EKM adds a switch. If you revoke your key, OpenAI typically can no longer read the workspace's stored data within an hour.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next