AI age verification and privacy: what 9 AI apps collect and how long they keep it

The Morse teamPublished November 4, 202610 min read

AI age verification costs you privacy twice: once when an app estimates your age from data it already holds, and again if it asks for a selfie or government ID. As of November 4, 2026, ChatGPT, Character.AI, Google, and Meta all run that estimate in the background, and OpenAI says its signals include the topics you chat about. A selfie or ID goes to an outside verifier such as Persona, Yoti, Incode, or k-ID, and among the apps that publish a period, it's kept anywhere from not at all, because it never leaves your phone, to 30 days.

We read each app's own help pages and privacy policy on November 4, 2026, and the policies of the verifiers they name, and put the answers in one table. Most of the top results we found cover ChatGPT alone and repeat OpenAI's seven days. Read side by side, the pages also show that age prediction runs on what you've already typed, searched, or posted, that Persona's own default is to delete at once, faster than the seven days OpenAI and Character.AI state, and that Yoti's own policy keeps an ID longer than Anthropic's summary of it says.

What do AI apps ask for to verify your age?

Most ask for nothing until they guess you're under 18 or a law applies, and then for a selfie, a government ID, or both. Each row is what the company's own pages say, read on November 4, 2026. Where a company has no age page for its AI app, we used the page for the account the app signs you in with, and say so.

Google's age checks cover every Google product, and Gemini needs a Google Account, at 13 or older for a personal one (Google). Meta's pages describe these checks for Facebook and Instagram, and Meta AI signs you in with a Meta account (Meta). We found no age page written for either AI app. We also looked for Talkie's privacy policy, but its link returned "Page not found" on November 4, 2026, so it isn't in the table.

Does ChatGPT need your ID?

No, not to use it. As of November 2026, OpenAI asks for a selfie or ID only when age prediction moves your account into ChatGPT for Teens and you want out, or where a law requires a check, and the account stays active either way (OpenAI). The table above has the methods and the deletion period, and our page on which AI chatbots need a phone number compares what 13 apps ask for at sign-up.

Italy is the exception. There, verification is required by a deadline, and an account deactivated for missing it can be restored through a separate check run by Stripe, a second verifier OpenAI names on its page (OpenAI).

Does AI age prediction read your chats?

At OpenAI, yes, in general terms. Its help page lists "general topics you talk about" among the signals. OpenAI gives two options if you'd rather it didn't predict: email privacy@openai.com to object, or verify with Persona, after which "OpenAI will not run age prediction on your account" (OpenAI).

The others are vaguer or look elsewhere. Character.AI's FAQ asks itself, "Is it looking at chats, characters, etc.?", and answers with "activity on the platform" without saying yes or no (Character.AI). Meta analyzes "entire profiles for contextual clues" in posts, comments, bios, and captions, and since May 2026 scans photos and videos for "visual clues about a person's age," which a July 31, 2026 correction says uses no biometric data (Meta).

Either way, the first step runs on what you've already typed, searched, or posted, and you send nothing for it. Our survey of what data AI chatbots collect covers the profiles these signals come from.

Does Character.AI's age verification need a selfie?

Only if its background check flags you as under 18 and you want the adult experience back, and we cover how its selfie and ID steps work in are Character.AI chats private. Its FAQ adds that if Character.AI launches "virtual currency or similar features," it may need "very limited exceptions where we work with Persona to access an ID" to resolve fraud, and would ask for your consent first (Character.AI).

What happens to your ID after age verification?

In most of these apps, the verifier deletes it on a schedule agreed with the app, and the app keeps only the result, such as a date of birth or a pass. The verifier's own policy and the app's summary of it don't always match.

• Persona's default is faster than its customers'. Persona's own policy, last updated April 9, 2026, says its default is to delete age-check data "immediately as soon as processing is complete," and that customers "may direct us to retain certain data" longer to detect or prevent fraud (Persona). OpenAI and Character.AI both state seven days, and Character.AI says Persona processes the data "to verify your age, identify fraud, and for quality assurance."
• Yoti keeps IDs longer than Anthropic's summary says. Anthropic's page says Yoti deletes your selfie and document images "as soon as your age is checked." Yoti's own privacy page says the same for most data, then adds: "If you added an ID Document we may keep this for 28 days for manual verification" (Yoti). A selfie-only check avoids that.
• Meta holds the selfie image itself. Instagram shares one image from your video selfie with its vendor, and "Once your age has been estimated and shared with us, we will delete the image within 30 days" (Instagram). For IDs, Facebook says Meta tells you the storage period before you upload, and that an ID may also be used to improve "our systems for detecting fake IDs" (Facebook).
• Google keeps the result in your account. An ID uploaded for age is "deleted after your date of birth is successfully verified," but that date of birth "will be saved to your payments profile," and Google may use the ID to "improve our verification services" (Google). For selfies, Google sends Private ID "only a random identifier for your Google Account."
• Replika and Janitor AI try the phone first. Replika's policy, updated October 28, 2026, says the selfie is analyzed on your device where possible and "never leaves your device," and goes to Incode only when that fails, to be destroyed within 24 hours (Replika). Janitor AI's k-ID selfie runs on your device, and k-ID's policy says its on-device estimate, provided by Privately, means images "are never uploaded to a remote server" (k-ID).

The check can also reach past the photo. Depending on the method, Persona's policy says it may confirm details against "government and national ID registries, consumer credit bureaus, utility companies, mobile network providers and postal address databases" (Persona). Replika's policy calls its deletion paragraph its "public written retention schedule," and lists what it keeps afterward: an age band, the method, and a timestamp. Our pages on whether Replika is private and whether PolyBuzz is safe cover those apps' other data.

Is an identity check the same as an age check?

No, and an identity check can keep your ID far longer. The same verifier can run under different rules:

• Claude has a separate identity check through Persona for "certain capabilities" and "routine platform integrity checks," which asks for a physical government photo ID and possibly a live selfie. Anthropic says it is "the data controller" and sets "how long it's kept," but the page, dated August 11, 2026, gives no period. Anthropic can view the records through Persona's platform, for example to review an appeal (Anthropic).
• Persona's policy has a separate section for identity verification. There, subject to the customer's retention period, it destroys ID and selfie scan data "upon completion of identity verification services or within three years of your last interaction with Persona" (Persona).
• Google saves an ID or selfie video given "to access additional services, features, or for other purposes" in your Google Account, where you can delete it from the IDs and Selfie video pages. For YouTube's advanced features, it's deleted automatically "within 2 years." If you've violated a Google policy, Google "may save your ID for a while." A separate, optional setting lets Google use your selfie video to develop "facial recognition, age estimation, and other verification methods" (Google).

Before you upload, read which check the prompt is asking for, since the two fall under different retention rules.

What do the laws say about AI age checks?

Some require the checks, and fewer say when the data must go. This isn't legal advice.

• United Kingdom. Section 12 of the Online Safety Act 2023 requires a service to use "age verification or age estimation (or both)" that is "highly effective" to keep children from the most harmful content (legislation.gov.uk). The section sets no deletion period. PolyBuzz cites the Act for its UK checks, and Janitor AI added the UK on June 15, 2026.
• New York. The SAFE for Kids Act says age data "shall be deleted immediately after an attempt to determine a covered user's age," though it covers social media feeds, not chatbots (NY Senate).
• Australia and Brazil. Janitor AI names Australia's age-restricted material codes and Brazil's Law No. 15.211/2025 as the reasons it checks users there from April 24, 2026 (Janitor AI).

How Morse keeps your AI chats from feeding an age profile

Every age guess in this article is built from an account the AI company holds on you: your chats, your searches, your posts, and how long you've been around. Every ID check ends by adding a date of birth or a verified flag to that same account. Morse is a web app with over 30 models, open and closed, built so the model makers never get your account and your chats don't sit on our servers. Morse is for adults only, and you confirm you're 18 or older.

You sign in with a code emailed to you, with no password, or with Google, and the Free plan needs no card. Each chat then runs in the privacy mode you pick:

• Anonymous, to use GPT, Claude, or Gemini without an account of yours at OpenAI, Anthropic, or Google. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider reads the question without knowing it's you. It may keep the message for a limited time to check for abuse, under its own policy, so leave your name and birthday out of what you type.
• Private, for chats you'd rather no host kept. It's the default for a new chat. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, run with your identity stripped, and Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing.
• E2EE, for the conversations no one should read, Morse included. Your browser checks the secure enclave's proof that it's genuine, then encrypts the message so only that enclave can open it, and our servers relay text they can't read. It runs open-weight models, rests on the enclave hardware and attestation working as designed, and comes with Pro, at $19 a month.

In every mode, your chat history is stored only in your browser, encrypted, and Morse's servers don't keep messages or log message content. Your Morse account holds your sign-in details, plan, balance, and a usage ledger with no chat content. Morse doesn't train AI on your chats, sell your data, or show ads (Morse privacy policy). Clearing your browser or deleting your account deletes your chats, and Morse can't recover them, so keep a copy of anything you need.

Send a first message from the Morse home page without an account, or start free with an emailed code and no card.