AI app privacy labels: what 16 chat apps declare and what they leave out

The Morse teamPublished November 6, 202610 min read

AI app privacy labels are the developers' own answers, and nobody checks them before they appear. On November 6, 2026, we read the labels of 16 AI chat apps on Apple's US App Store: 10 listed the content you type, 5 declared data used to track you, and 4 listed nothing you type although their own privacy policies say they collect your messages. Those four are Perplexity, Replika, Pi, and Kimi.

We read each app's "App Privacy" section on apps.apple.com, then its "Data safety" section on Google Play where we could find one (15 of the 16), then the privacy policy each label links to, all on November 6, 2026. We counted Apple's top-level categories (Apple has 16, from Contact Info to Other Data) in each section of each label. Perplexity's site blocked our requests, so its policy is quoted from an archived copy. Venice's and Lumo's short labels match what their policies describe.

What do AI app privacy labels say?

Every label sorts data into up to three sections: Data Used to Track You, Data Linked to You, and Data Not Linked to You. The table lists Apple's categories in each, as each label showed them on November 6, 2026, with the developer named on the label. "Lists what you type" means User Content appears in any section; for chat apps, Apple's developer guidance says free-form text fields go under "Other User Content."

Three labels list Health & Fitness as linked to you: ChatGPT, Gemini, and Meta AI. Two list Sensitive Info, which Apple defines to include sexual orientation, religious beliefs, and biometric data: Gemini and Meta AI. Grok's and Meta AI's also list "Emails or Text Messages" among the user content linked to you.

What does "Data Linked to You" mean on ChatGPT?

It means OpenAI says it may collect eight kinds of data from the app and tie them to your account: Health & Fitness, Location, Contact Info, User Content, Search History, Identifiers, Usage Data, and Diagnostics. Apple counts data as linked unless it's stripped of identifiers before collection and never tied back to you.

The label's detail view, opened from "See Details" on ChatGPT's App Store page, shows the purpose for each. On November 6, 2026:

• Third-Party Advertising: Usage Data, listed as "Product Interaction" and "Advertising Data." Apple's definition of the purpose includes "displaying third-party ads in your app."
• Developer's Advertising or Marketing: your email address and Usage Data.
• Analytics: Health & Fitness, Contact Info, User Content, Identifiers, and Usage Data.
• Product Personalization: Health & Fitness, your name, and Usage Data.
• App Functionality: coarse location, your name, email, and phone number, audio, customer support messages, other user content, search history, a user ID and a device ID, usage data, and crash and performance data.

ChatGPT's label declares nothing used to track you. Its Google Play listing says it shares "Device or other IDs" for "Advertising or marketing" and for fraud prevention, and marks your messages and "Other user-generated content" as optional. The Play listing has no health data at all, while the Apple label lists Health and Fitness for analytics and personalization. Both labels list only coarse or approximate location, though the iOS and Android apps can share your device's precise location. OpenAI says ChatGPT "does not separately save or store your precise location outside of what information is included in your ChatGPT conversation and context" (OpenAI), and how ChatGPT knows your location covers the rest. OpenAI's privacy policy is covered in what data AI chatbots collect.

Which AI apps declare tracking on the App Store?

Five of the 16: Grok, Character.AI, Replika, Copilot, and Kimi. Apple defines tracking as linking data from your app "with Third-Party Data for targeted advertising or advertising measurement purposes," or sharing it with a data broker, and its developer page adds that "Third-Party Data" is data from "apps, websites, or offline properties not owned by you."

Meta AI's label declares nothing used to track you, yet it lists Third-Party Advertising as a purpose for 12 categories linked to you, among them Browsing History, Contacts, Financial Info, and Location. Under that definition, data from Facebook, Instagram, and Meta's other apps isn't third-party data to Meta, since Meta owns them. Five labels list Third-Party Advertising as a purpose: Meta AI, ChatGPT, Copilot, Character.AI, and Replika. Two of the five, Meta AI and ChatGPT, declare no tracking. Which trackers load inside these apps is a separate question, covered in do AI chatbots share your data with trackers.

Do AI app privacy labels match the privacy policies?

Not for four of the apps that leave out what you type. Apple's definition of "collect" is "transmitting data off the device in a way that allows you and/or your third-party partners to access it for a period longer than what is necessary to service the transmitted request in real time." A chat app that keeps your messages after answering, or uses them to train its models, meets it. Here is what each policy said on November 6, 2026:

• Perplexity. Its label lists no User Content and no Search History. Its privacy notice, last updated October 19, 2026, says "We process content you submit while using our services, including prompts, queries, uploads," and that accounts can "save query history" (Perplexity). We read an archived copy from October 23. The same notice says it may collect health data if you enroll in Perplexity Health; the label lists no Health & Fitness. On location, the label says more than the notice. It lists Precise Location for product personalization, while the notice describes only "general location from the IP address," as we noted in what data AI chatbots collect. More on Perplexity in is Perplexity private.
• Replika. Its label lists no User Content. Its policy, updated October 28, 2026, collects "the messages you send and receive through the Apps" and keeps "your conversation history for 24 months for text messages" (Replika). The label's tracking section lists Identifiers and Usage Data. The policy also describes sending Meta and Google a hashed email address outside the EEA, the UK, Brazil, and Switzerland, which "our partners use … to match your activity to an account you may have with them," plus purchase events "including the amount you paid." Apple's examples of tracking include sharing "a list of emails" with an ad network. We cover the rest in is Replika private.
• Pi. Its label lists Identifiers, Usage Data, and Diagnostics. Inflection's policy collects "inputs" of "text, voice, audio, or other materials," keeps them for "at most 15 days after user deletion," and keeps outputs "indefinitely" (Inflection).
• Kimi. Its label lists only a device identifier. Moonshot's international policy, updated November 5, 2026, covers "prompts, audio, images, videos, files, and any content you input or generate," used for "training and optimizing" its models (Kimi). The label's privacy link leads to Moonshot's mainland China policy, as is Kimi AI safe explains.

Two short labels do match their policies. Venice says it "will not have access to or store the content of your Prompts or Outputs" (Venice), and Proton says saved Lumo chats are encrypted "with zero-access encryption (meaning we as a provider are not able to decrypt it anymore)" (Proton). Under Apple's definition, neither collects your chats.

The developer's name can differ too. Grok's Apple label names X Corp. The policy it links to belongs to SpaceX Super Intelligence LLC, last updated October 27, 2026, which says "SpaceXSI is a separate company from X Corp." (SpaceXSI). Grok's Play listing names a third company, SpaceXAI LLC.

Why do App Store and Google Play labels disagree?

Each store asks its own questions, and the same developer fills in each form. On November 6, 2026, Google Play's Data safety sections differed from the Apple labels in several ways:

Part of the gap is in the definitions. Google says "'Sharing' refers to transferring user data collected from your app to a third party," and leaves out transfers to service providers, for legal reasons, at a user's request, or of fully anonymized data (Google). Apple's tracking section asks only whether your data is combined with other companies' data for ads or sent to a data broker. On Google Play, "'Optional' includes the ability to opt into or opt out of data collection," and ChatGPT, Meta AI, Grok, Poe, and Kimi all mark the content you type as optional.

Who checks AI app privacy labels?

Nobody, before you read them. Every one of the 16 Apple labels says "This information has not been verified by Apple." Apple tells developers: "You're responsible for keeping your responses accurate and up to date," and lets them change their answers "at any time" without an app update (Apple). Google tells developers: "You alone are responsible for making complete and accurate declarations," and "Google's review process is not designed to verify the accuracy and completeness of your data safety declarations" (Google).

A label describes the app only, as of the day you look. To check one, find User Content in any of its sections, compare the Apple and Google versions, and open the privacy policy it links to. The policies are long too. The median AI privacy policy took about 18 minutes to read when we counted their words.

How Morse keeps your chats off its servers

Every label above is a company's own account of what it keeps on its servers and ties to your name, and neither store checks it. Morse is built so your chats aren't on its servers in the first place. It's a web app at trymorse.ai, so it has no App Store or Google Play listing; what it keeps is in the Morse privacy policy.

Your chat history is stored in your browser, encrypted, not on Morse's servers. Our servers don't keep your messages or log what you write, and your account holds only your sign-in details, plan, balance, and a usage ledger with no content. Morse runs no advertising trackers or third-party analytics, shows no ads, doesn't sell your data, and doesn't train AI on your content. You pick a privacy mode for each chat:

• Private, so no model host keeps a copy of what you type. New chats start here. 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. Open-weight models like DeepSeek, Kimi, and GLM run here, and so does Grok.
• Anonymous, so GPT, Claude, and Gemini get your message without knowing it's you. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider may keep it for a limited time to check for abuse, under its own policy, so leave out details that point to you.
• E2EE, for anything no one else should read. Your browser checks the enclave's hardware attestation before anything leaves your device, then encrypts the message so only that enclave can read it, and our servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed.

In every mode, your history is locked when you sign out and takes two keys to open, one of which never leaves your device. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you'll need.

Send a first message from the Morse home page without an account, or start free with no card. End-to-end encrypted chats come with Pro, at $19 a month.