AI privacy for LGBTQ people: who could learn who you are from a chat, and what stops them

The Morse teamPublished November 4, 202610 min read

AI privacy for LGBTQ people depends on who else can reach your chats: someone who opens a device or account you're signed in to, an employer that runs the account, the AI company and the people who review its chats, advertisers in apps that show ads, a government that asks for records, or whoever gets in through a breach or a shared link. As of November 2026, OpenAI, Google, and Meta keep what you type into ChatGPT, Gemini, and Meta AI on their servers, so a question about coming out or transitioning can reach any of those audiences unless you pick a tool that stops it.

We read GLAAD's 2026 AI report, ILGA World's Pride Month 2026 data on laws affecting LGBTI people, EFF's Surveillance Self-Defense guide for LGBTQ youth, the GDPR, California's privacy regulator, and OpenAI's, Meta's, and Google's own ad and privacy pages on November 4, 2026. This page is about privacy, not legal advice.

GLAAD's report, Build for Everyone, says LGBTQ people "face heightened risks when AI systems collect, infer, or retain information about sexual orientation, gender identity, or other personal characteristics." Each risk below has its own fix, and most of them don't involve leaving AI behind.

Is it safe to ask ChatGPT about gender identity?

It can be, if you control where the copy goes. A question about pronouns, hormones, or coming out is saved to your account like any other chat, and the companies say plainly what that copy is open to:

• Human review. Google's Gemini privacy hub says "a subset of chats are reviewed by human reviewers," that reviewed chats "are retained for up to three years," and asks: "Please don't enter confidential information that you wouldn't want a reviewer to see" (Google). We compare the reviewer rules at each company in who reviews AI chats.
• Retention after you delete. Meta's AI terms, effective May 13, 2026, say "Do not share information that you don't want the AIs to use and retain, such as information about sensitive topics," and warn that deleting messages, threads, "or even your account may not delete our copy of your personal information" (Meta).
• Training. OpenAI's US privacy policy lists "using your Content to train our models" among its uses, subject to your control in Data controls (OpenAI).

GLAAD's answer for companies is to "minimize the sensitive data they collect, infer, and retain." Until they do, the person asking does the minimizing: leave out your name, your town, and anyone else's name, and pick a chat that isn't saved, or one the company can't read.

Can ChatGPT out you?

We found no report of a chatbot telling someone else a user's sexual orientation. The outing happens around the chatbot, in the way EFF's guide for LGBTQ youth describes for the web in general: "accidental outing to your peers, family, or online advertisers as a result of online tracking or nosy snoopers." With AI chats, it comes four ways.

• Someone opens your account. Your chats belong to the account, so they show on every phone, laptop, or tablet signed in to it. ChatGPT's Locked Chats hide chosen conversations behind a 4-digit PIN, and OpenAI says they help "when you share your screen or use a shared device." But OpenAI also says locking a chat does not "create end-to-end encryption" and does not "remove information that ChatGPT remembered before you locked the chat" (OpenAI). If memory is on, check what it holds; how to turn off ChatGPT memory shows where to look.
• An ad on a shared screen. GLAAD's report says that when platforms use inferred LGBTQ identity data to serve ads, "they could inadvertently 'out' users to others who share devices or see their screens."
• A link you shared. OpenAI says locking a chat that already has a shared link "does not delete the link" (OpenAI). We cover how shared chats ended up in search results in AI chat share links.
• A bug. On March 20, 2023, a bug in an open-source library "allowed some users to see titles from another active user's chat history" (OpenAI). A title is often the question itself.

For a partner or ex who checks your devices on purpose, the steps are different, and some of them can alert the other person. Our guide for domestic violence survivors covers checking signed-in sessions quietly.

Are AI chats about sexuality used for ads?

It depends on the app. Meta, Google, and OpenAI each draw the line in a different place.

• Meta has used Meta AI chats to pick ads and posts since December 16, 2025, with no opt-out described in its announcement. It names sexual orientation among the topics it leaves out: when people talk with Meta AI about "religious views, sexual orientation, political views, health, racial or ethnic origin, philosophical beliefs, or trade union membership," Meta says "we don't use those topics to show them ads" (Meta).
• Google says "Your Gemini Apps chats are not being used to show you ads" (Google). Its ad policy lists "Sexual orientation" and "Transgender identification" as sensitive interest categories, and advertisers in them can't use their own curated audiences (Google).
• OpenAI shows ads on ChatGPT's Free and Go plans. With personalization on, it may use "Past chats and memory" to choose them, and it says "Advertisers never receive your chats, chat history, memories, name, email, precise location, IP address, or sensitive information (such as health, mental health, political topics)" (OpenAI).

OpenAI's pages are the only ones of the three that don't name sexual orientation or gender identity among the chats kept away from ads. Its ad policy's placement rules bar ads near "Mental and personal health conversations," "Debated social content," and "Sensitive user journeys" (OpenAI), and OpenAI doesn't say whether a chat about coming out falls under any of them. In that policy, gender identity appears only in the rules for the ads themselves, which may not "meaningfully reference sensitive social issues," such as "abortion, immigration, race, religion, or gender identity." Since April 2026, the same policy says medical, legal, and financial advice contexts "are no longer categorically blocked from ads by default."

Can your employer see your AI chats?

Yes, on a work account. OpenAI's privacy policy says administrators of a ChatGPT Enterprise or business account "may access and control your OpenAI account, including being able to access your Content." It adds that if you sign up with an email address belonging to your employer, OpenAI may share "the fact that you have an account" and your email address with that employer (OpenAI). Keep personal questions on a personal account, a personal email, and a device your employer doesn't manage; AI on a work laptop covers what the device itself reports.

Is sexual orientation protected data under privacy law?

In Europe and California, yes, with limits. Article 9(1) of the GDPR says the processing of "data concerning a natural person's sex life or sexual orientation shall be prohibited," unless an exception in Article 9(2), such as explicit consent, applies (EUR-Lex). The article doesn't name gender identity. California's privacy regulator counts "information about a consumer's health, sex life, sexual orientation" as sensitive personal information, and says you can direct businesses to use it only for limited purposes, "such as providing you with the services you requested" (CPPA).

These rules limit what a company may do with your chats, not whether a court can reach them. Article 9(2) itself makes an exception for legal claims and "whenever courts are acting in their judicial capacity." This isn't legal advice.

What about countries where being LGBTQ is a crime?

There, a stored chat can become evidence. ILGA World's Pride Month 2026 data counts 65 UN member states that criminalise consensual same-sex sexual acts, and says the total "grew for the first time in almost a decade" after Burkina Faso passed such a law and Trinidad and Tobago reversed its decriminalisation. The death penalty is the prescribed penalty in 7, with 5 more lacking full legal certainty, and at least 63 limit freedom of expression related to sexual, gender, and bodily diversity. Kazakhstan and Belarus adopted laws banning information on "non-traditional sexual orientations" and "sex change," and Malaysia blocked the websites of several dating apps linked to the community (ILGA World).

GLAAD's report cites "more than 60 countries," where "government access to AI-collected information can lead to arrest, persecution, or violence." It adds that in the many US jurisdictions that restrict transgender rights, the same data "can fuel discrimination, denial of care, or loss of legal recognition."

Access comes two ways. A government can ask the company: OpenAI's privacy policy says it may share personal data with government authorities "if required to do so to comply with a legal obligation" (OpenAI), and our page on whether AI chats can be subpoenaed shows how often companies are asked. Or it can take the phone, which our guide to AI chats at the border covers for US crossings.

Who can learn what from your AI chats, and what helps

We matched each audience to what it can see and the fixes that work, from the pages above, as of November 4, 2026.

How Morse keeps questions about who you are off servers and other people's screens

Nearly every row in that table starts with the same thing: a copy of your chats on a company's servers, tied to an account that opens on any device you've signed in to. Morse is built without that copy. Our servers don't keep your messages or log what you write, and your chat history is stored only in your browser, encrypted.

Take the table's rows in order. On a shared family computer, history is locked when you sign out, and unlocking it takes two keys, one that never leaves your device and one our server gives you at sign-in, so a copy of the browser's data alone can't be read. On a computer that isn't yours, a message sent from the Morse home page opens a guest chat that's kept in the tab's memory and never saved. You sign in with a code sent to your email or with Google, so use an address your employer and family can't open. Morse doesn't show ads, sell your data, or train AI on your content, so there's no ad to carry a chat onto a shared screen. Because we don't store chats, we usually have very little to give authorities (privacy policy).

Each chat picks how far the message itself travels:

• E2EE, for what you haven't told anyone yet. Use it for a letter to your parents, questions about transition, or anything you'd never want a company to hold. Your browser checks the secure enclave's proof that it's genuine before anything leaves your device, then encrypts the message so only that enclave can read it, and our servers relay text 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. E2EE chats come with Pro, at $19 a month.
• Private, for everyday questions about dating, health, or your rights. It's the default for a new chat. Morse has zero data retention agreements with its inference, search, and other providers, so the hosts running open-weight models like DeepSeek, Kimi, and GLM keep nothing.
• Anonymous, for asking GPT, Claude, or Gemini without your name on it. Our proxy removes your account, email, and IP address, so the provider sees 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 out names and places.

Clearing your browser or losing the device deletes your history for good, and we can't recover it, so keep a copy of anything you'll want later somewhere only you can reach.

Try a guest chat from the Morse home page without an account, or start free with no card.