Can the WiFi owner see your ChatGPT chats? What a network sees, and what it can't

The Morse teamPublished November 3, 20269 min read

A WiFi owner can't read what you type into ChatGPT, but they can see that you used ChatGPT, when, and for how long. OpenAI says your content is "encrypted at rest and in transit between you and OpenAI" (OpenAI), so a home router, a café hotspot, or your internet provider sees the name chatgpt.com and the size and timing of the traffic, not your prompts or the replies. The exceptions are a school or work network that decrypts traffic with a certificate installed on your device, and monitoring software on the device itself.

We read the internet standards that define what a connection reveals, the FTC's study of internet providers, Apple's, Cloudflare's, and OpenAI's own documentation, two school and workplace filtering vendors' help pages, and two published attacks on encrypted AI chat traffic, all on November 3, 2026. We also looked up each major AI chat site's public DNS records that day, to see which ones can hide their name from the network. The FTC deserves credit for saying plainly, in its public WiFi guidance dated February 2023, that "connecting through a public Wi-Fi network is usually safe" because of "the widespread use of encryption" (FTC).

What can a WiFi owner see when you use ChatGPT?

The site's name, and when. Before your browser can reach ChatGPT, it looks up chatgpt.com in DNS and then opens an encrypted connection, and both steps usually carry the name in the clear. The IETF's own document on the problem says the Server Name Indication, the field that names the site you're connecting to, "is carried in cleartext in the TLS 'ClientHello' message," and lists who reads it: surveillance, content filtering by ISPs for parental controls, and "firewalls within enterprise networks blocking websites not deemed appropriate for work" (RFC 8744).

A home router shows the owner some of this. NETGEAR's activity log, for one, is "a detailed record of the websites you have accessed or attempted to access," listing the date and time, the device's address on the network, and the site's name. It keeps up to 256 entries, and it only records them when keyword blocking is on (NETGEAR, updated July 7, 2025). Nothing in it shows what you asked.

Who on the network sees what, as of November 2026

No network setting changes what OpenAI receives. Its privacy policy says its log data "includes your Internet Protocol address, browser type and settings, the date and time of your request," and that it works out "the general area from which your device accesses our Services" from that address (OpenAI, read November 3, 2026).

Can my internet provider see what I ask ChatGPT?

No, but it sees where you go. The FTC's 2021 staff report on six major providers, AT&T, Charter, Comcast, Google Fiber, T-Mobile, and Verizon, found that "many of the ISPs in our study have access to 100% of consumers' unencrypted internet traffic," and that "even with encryption, many ISPs in our study continue to store the IP addresses that their customers access and thereby collect the domain names of the websites they visit" (FTC, October 21, 2021). Two of those providers told the FTC they use web browsing information to target ads. ChatGPT is encrypted, so for a chat that means a record of the visit, not its content.

Does encrypted DNS or Encrypted Client Hello hide ChatGPT from the network?

Encrypted DNS hides the lookup, and Encrypted Client Hello would hide the name in the handshake, but no major AI chat site uses the second one yet. DNS over HTTPS "encrypts DNS queries by wrapping them inside regular HTTPS requests" (Cloudflare, updated May 5, 2026). On its own, that still leaves the site's name in the handshake.

Encrypted Client Hello (ECH) closes that gap. It became an internet standard in March 2026, and the standard lists its limits. Using it "reveals that a client is connecting to a particular service provider," and the site "may also be visible through other channels, such as plaintext client DNS queries" (RFC 9849). Cloudflare, whose network serves several of the sites below, turns it on by default only for Free plans; "other plans can turn it on" themselves (Cloudflare, updated April 17, 2026).

A browser can only use ECH when the site publishes a key for it in DNS. We looked up the HTTPS records of 13 AI chat sites on November 3, 2026: chatgpt.com, claude.ai, gemini.google.com, grok.com, www.perplexity.ai, chat.deepseek.com, copilot.microsoft.com, www.meta.ai, character.ai, duck.ai, lumo.proton.me, chat.mistral.ai, and trymorse.ai. None of them published an ECH key. So today, on every one of them, the site's name crosses the network in the clear, whatever your browser's DNS setting.

Can a school or work network read my ChatGPT chats?

Only by decrypting your traffic, and that needs its certificate on your device. School filters and corporate gateways do this with TLS inspection. Securly, a school web filter, says installing its certificate "ensures that Securly Filter can filter all HTTPS sites," and calls deploying it through device management "the preferred deployment method as it ensures all managed devices receive the trusted SSL certificate." For students' own devices, it offers a download page (Securly). Zscaler describes its service as "a full SSL/TLS proxy, or SSL/TLS man-in-the-middle (MITM) proxy" (Zscaler).

Inspection tends to show itself with ChatGPT. OpenAI's help center warns that "SSL inspection / decryption on your network may lead to SSL errors and disrupt access to the app" (OpenAI). If you ever installed a school or employer certificate on a personal phone to get online, that network can read what the phone sends through it.

A managed laptop is a separate problem from the network, because software on the device can capture a prompt before it's encrypted. We cover that layer in can my employer see my ChatGPT and, for school accounts and Chromebooks, in AI privacy for students.

Can someone on the same public WiFi read my AI chats?

Not the words, but researchers have shown they can sometimes guess the subject. In March 2024, researchers at Ben-Gurion University showed that the sizes of encrypted packets in a streamed reply could leak the length of each word-piece, and reported they could reconstruct 29% of an AI assistant's responses and infer the topic of 55%, with ChatGPT-4 and Microsoft Copilot among the services tested (arXiv). Cloudflare published a fix the same day, padding its streamed AI responses so packet sizes no longer reveal token lengths (Cloudflare).

Microsoft found a second leak in November 2025 and called it Whisper Leak. It uses "network packet sizes and timings" to tell whether a chat is about a chosen topic, by someone such as "a nation-state actor at the internet service provider layer" or "someone connected to the same Wi-Fi router." It doesn't recover the text. Microsoft reports that OpenAI and Microsoft Azure added an "obfuscation" field of random text to streamed responses, Mistral added a similar parameter, and xAI deployed protections (Microsoft, November 7, 2025). Its advice for users includes avoiding "highly sensitive topics" on untrusted networks and using a VPN.

Both are research attacks, run in lab conditions.

Does a VPN or iCloud Private Relay hide ChatGPT from the WiFi owner?

Yes, from the WiFi owner, by moving what they'd see to someone else. With a VPN, the network sees a connection to the VPN, and the VPN provider sees the site names instead. Apple's iCloud Private Relay splits that view between two relays so that no single party, Apple included, can see both who you are and which sites you visit, but it covers browsing in Safari and comes with an iCloud+ subscription (Apple). Networks "that require the ability to audit traffic or perform network-based filtering" can block it, Apple says, which is why it may stop working on a school or office network (Apple, October 1, 2026).

Neither one changes what OpenAI receives. We compare VPNs with every other way of hiding from OpenAI in how to use ChatGPT anonymously and in AI privacy myths.

How Morse keeps your chats private on any network

Every network above sees the same thing: a site's name and the shape of the traffic. The words themselves arrive at the company on the other end, which keeps them under your account, so the choice of AI service settles far more than the choice of WiFi. Morse is built so that company doesn't learn who you are, and so your chat history stays on your device instead of under an account.

• Anonymous, to use GPT, Claude, or Gemini without the provider knowing it's you. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider still sees the message and may keep it for a limited time to check for abuse, under its own policy, so leave out names and details you'd rather keep to yourself.
• Private, for everyday chats no host should keep. New chats start here, with open-weight models like DeepSeek, Kimi, and GLM, plus Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for anything no one should read. Your browser checks the secure enclave's proof that it's genuine, then encrypts the message so only that enclave can read it. Morse's servers relay ciphertext they can't read, and the reply is decrypted only in your browser. This rests on the enclave hardware and attestation working as designed, and the enclaves are run by third-party providers, not by Morse.

In every mode, your chat history is stored only in your browser, encrypted, not on Morse's servers. Morse doesn't train AI on your chats, sell your data, or show ads, and because we don't store chats, we usually have very little to give authorities.

Morse can't hide its own name from the network. As of November 3, 2026, trymorse.ai publishes no ECH key either, so a WiFi owner sees that you visited trymorse.ai, the way they'd see chatgpt.com. If even that matters, use your own phone's mobile data, and keep personal chats off any device your school or employer manages.

So who can see your AI chats on WiFi?

• A home router, a landlord, or a café: the site's name and the time. Not your prompts, and not your history.
• Your internet provider: the same, kept longer, and possibly used for ads. Encrypted DNS hides the lookup; nothing yet hides the name in the handshake on the major AI sites.
• A school or work network: possibly everything, if your device trusts its certificate. Use a device and connection that are yours.
• Someone on the same public WiFi: the site's name and, in research conditions, a guess at the topic. On an untrusted network, keep the most sensitive chats for later or use a VPN.
• The AI company: every word, tied to you. On Morse, Anonymous strips who you are, Private keeps nothing at the host, and E2EE leaves our servers with only ciphertext.

You can try a chat from the Morse home page without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.