Is Chai AI safe? What Chai keeps, what it trains on, and who it's for

The Morse teamPublished October 28, 20268 min read

As of October 2026, Chai AI is safe for adults who are comfortable with a company keeping their conversations: Chai Research says it uses your chats with its bots to fine-tune its models, with personal identifiers removed first, and may keep your data for up to five years after you delete your account. It's not for anyone under 18, and Chai now checks ages and shows crisis resources when a message talks about self-harm.

We read Chai's privacy policy (last updated October 12, 2026), its end user license agreement (July 28, 2026), its safety center, its company facts, and its App Store and Google Play listings on October 22, 2026, along with the original reporting on the 2023 case in Belgium. Chai earns credit for a public safety center, age checks run by an outside identity company, and a company-facts page its CEO reviews every month.

What Chai does with your chats, and what you can do about it

Chai Research Corp., a Delaware company based in Palo Alto, has run Chai since January 2021 (Chai). Each row below is one worry, what Chai's own page says about it on October 22, 2026, and the step left to you.

Are Chai AI chats private?

From other users, yes, and from Chai, no. Chai's privacy policy says your personal data "will be accessible only by specifically authorized staff," and that it is also shared with Chai's suppliers, which act "as data processors" on Chai's instructions. The policy doesn't name those suppliers. It adds that your data "may also be disclosed to third parties to fulfill legal obligations and/or requests made by competent authorities."

Chai also keeps a record of what you do. "To resolve technical malfunctions, we log activities associated with your user ID," the policy says, giving "login times, chatbot openings, message submissions" as examples. You sign in through Google, Facebook, or Apple, so that user ID sits next to an email address.

Chai's safety center says "Every user message is scanned in real time for self-harm content."

Does Chai train on your conversations?

Yes. Under the heading "Model improvement," Chai's privacy policy says: "We use personal data derived from your in-app conversations with chatbots to enhance and fine tune" its AI models. "Before we use any conversation data in this way, we remove personal identifiers." The policy doesn't list which identifiers it removes, and removing them leaves the story itself, so a chat that names your town or your employer still names them.

Chai's policy describes no training switch. It relies on "legitimate interest" as the legal basis, which gives people in the EU and UK a right to object "on grounds relating to your particular situation." Chai can refuse if it has compelling grounds of its own.

Chai's license agreement goes further for what you post. You keep ownership of your "Contributions," but you give Chai an "unrestricted, unlimited, irrevocable, perpetual, non-exclusive, transferable, royalty-free, fully-paid, worldwide right" to use them, including for "commercial advertising." We compare training rules across AI companies in which AI companies train on your data.

Why does Chai's Google Play privacy link open a 2021 page under another name?

As of October 22, 2026, the Privacy Policy link on Chai's Google Play listing opens a notice hosted on GitLab, last updated March 14, 2021, from "Nexus, doing business as chai" (notice). Chai's own pages don't explain the name. The App Store listing links to chai.ml/privacy, which now redirects to Chai Research's current policy. An Android user who checks the store link reads rules that differ from the ones Chai applies today:

The current policy's five years can stretch further where Chai needs the data "to defend and enforce its rights" (Chai), so the Delete Account button in the app's settings ends your account well before it ends Chai's copy.

The two guides in Google's top results for "is chai ai safe" that we read on October 22, 2026, from Parental-Control.net and FamiSafe, mention neither the five-year window nor the older notice. FamiSafe's guide, updated April 30, 2026, also gives Google Play's 17+ rating without noting that Chai's own terms set the minimum at 18.

Does Chai sell your data or show ads?

Chai's current policy doesn't use the word "sell." It says Chai uses personal data "to send you advertisements if you have a free subscription, based on your consent," and that if you installed the app through an ad, "we may also share certain information, such as your usage patterns, with our advertising partners" (Chai).

The store labels add detail. The App Store label says Usage Data is "Used to Track You" (App Store), and Google Play's data safety section says Chai shares approximate location and the list of installed apps for "Advertising or marketing" (Google Play).

Chai's wording ties the ads to the free tier. As of October 22, 2026, the App Store lists Premium at $13.99 a month or $134.99 a year, Ultra at $19.99 a week or $29.99 a month, and packs of Pieces from $2.79. Chai says that "in some regions, use of the service requires a paid subscription" (Chai). Its CEO, William Beauchamp, told Reuters on October 14, 2026, that Chai made "a big strategic push to minimize any user whose cost is greater than their revenue" (Reuters). Chai's company facts put its revenue at roughly 60% subscriptions and purchases and 40% advertising.

Is Chai safe for kids?

No, and Chai says so. Its safety center says "CHAI is strictly for adults," and its license agreement restricts it to users aged 18 and older. Age checks run through Persona, either "a quick selfie" to estimate whether you're over 18 or a government-issued photo ID. The App Store rates the app 18+. Google Play's US listing shows "Mature 17+." Adults can turn on filters that hide explicit messages and block characters, and Chai says its models are designed to refuse sexually explicit content.

What happened with Chai in Belgium in 2023?

On March 28, 2023, the Belgian newspaper La Libre reported that a man in his thirties had died by suicide after six weeks of conversations with a chatbot called Eliza, and published his widow's account (La Libre). The same day, La Libre reported that the chatbot's founder had sent it a screenshot of a new message pointing people with suicidal thoughts to help, one that hadn't appeared during the man's conversations (La Libre). Vice reported on March 30 that the bot ran on Chai, and quoted Beauchamp: "we worked around the clock to get this feature implemented" (Vice).

In August 2023, La Libre reported that a group called the Safe AI Companion Collective had filed complaints against Chai Research with Belgium's data protection authority and its federal economy service, over transparency, data protection, and age checks. The same article says no direct link between the man's death and the AI has been proven (La Libre).

Today Chai says that when a user "expresses self-harm intent," the app shows "a notification referring the user to crisis service providers" (Chai). If you or someone you know is struggling, you can call or text 988 in the US (988 Lifeline), or find a free helpline in your country through Find a Helpline.

How Morse leaves no five-year copy of your character chats

Every answer above comes back to Chai holding your conversations on its servers in the United States (privacy policy), tied to a Google, Facebook, or Apple login, for up to five years after you leave. On Morse, Characters, for roleplay and companion chats, keeps your character chats, the characters you make, and your personas encrypted in your browser. Our servers don't keep messages or log their content; your account holds sign-in details, your plan, your balance, and a usage ledger with no content in it (privacy policy).

We don't train AI on your content, so there's no legitimate-interest objection to file. Morse's servers keep no copy of your chats, so clearing your browser or deleting your account removes them for good, and we can't recover them either, so keep a copy of any story you want. Morse shows no ads and runs no advertising trackers or third-party analytics, and because we don't store chats, we usually have very little to give authorities.

Chai added its crisis notice after the Belgium case, and Morse has one too. Morse is for adults only, and if something you write in a character chat reads like suicidal thinking, Morse shows crisis lines, with your own device making that check.

Morse runs in your browser, and you sign in with a code sent to your email or with Google. The Free plan has no cap on character messages, with the model picked for you. Pro, $19 a month, lets you run character chats on any included model and adds unfiltered models.

In Chat, you pick a privacy mode for each conversation. In E2EE, on Pro and Max, your browser checks the enclave's hardware attestation, then encrypts the message so only that enclave can read it, as long as the enclave hardware and attestation work as designed. Private mode runs open-weight models such as DeepSeek, Kimi, and GLM under zero data retention agreements with Morse's inference, search, and other providers. Anonymous mode sends a GPT, Claude, or Gemini message under Morse's own account with your email and IP address removed, though that provider may keep it for a limited time to check for abuse.

For how other apps handle the same questions, see our comparison of Character.AI alternatives and the hub on whether AI chats are private.

Start free: the Free plan has no cap on character messages.