safety.

What Blank is, and isn’t

Blank is an iPhone app for practicing everyday conversations. Pip, an AI coach, sends one short conversation a day. An AI character plays the other person, and when the conversation ends, Pip gives feedback on what you said.

Blank is practice, not therapy. It isn’t a crisis service, a medical or mental-health service, or a substitute for professional care, and nobody watches conversations as they happen.

How the app spots possible crisis language

Every message you send in a practice conversation is checked in two ways:

• An automated check. Each message is read in the context of the scene and flagged if it suggests you may be thinking about suicide or self-harm. Everyday expressions like “I’m dying to leave” don’t count.
• Fixed phrase matching. Each message is also compared with a list of clear phrases about suicide and self-harm, so a plain statement is caught even if the automated check misses it.

Spoken replies are turned into text on your phone and checked the same way. No automated check is perfect: it can miss things, and it can occasionally pause a scene that didn’t need pausing. We’d rather pause by mistake than miss someone who needs help.

What happens next

If a message suggests you may be in crisis:

1. Practice pauses. The scene stops right away.
2. You see crisis lines. A card lists them: in the US, call or text 988, or text HOME to 741741 (Crisis Text Line); elsewhere, findahelpline.com. It also says: if you are in immediate danger, call your local emergency number. The card’s wording is fixed. It is never written by the AI.
3. Role-play is paused for 24 hours. You can still open the app, and the crisis resources stay in Profile → Help.
4. Nobody is contacted on your behalf. Blank doesn’t call emergency services, and it doesn’t contact anyone you know.

If a message is worrying but less clear, Pip pauses the scene and asks you directly whether you’re having thoughts of hurting yourself. If you say yes, you see the crisis lines. If you say no, the scene picks up where it left off.

Where scenes stop

Characters keep every scene PG. If a conversation turns to suicide or self-harm, the scene stops rather than continuing, as described above. If a message is sexually explicit, or contains slurs or threats of violence, the scene ends.

You’re always talking with AI

Pip and every character are AI, never a person. The app says so before your first conversation, where it also explains that conversations run on AI from Anthropic, and each conversation opens with a card that labels its character as an “AI character”. Characters stay in character so the practice feels real, but if you sincerely ask one whether it’s AI, it tells you plainly that it is. A character never claims to be human.

Once Blank has been open on your screen for 3 hours in total, it reminds you that you’re talking with AI while you’re in a conversation, and again after every 3 more hours. Time with the app closed or in the background doesn’t count.

For adults only

Blank is for adults 18 and over, and you confirm your age before your first conversation. If someone tells a character they’re under 18, practice stops and the app shows how to delete the account.

Notice under California Business and Professions Code §22604: AI chat, including Blank’s characters, may not be suitable for some minors.

Reports

You can report any message from a character in the app, or write to us through the support form. A person reviews every report within 2 business days and acts on what we find, for example by changing or removing a conversation.

What we count

Each time the app shows a crisis card, we record that it happened. After 90 days, the record is stripped of anything that could identify you and kept only as a count. Once a year, starting in 2027, we report the number of crisis referrals to California’s Office of Suicide Prevention, as Business and Professions Code §22603 requires. The report contains no personal information.

Changes and questions

We update this page whenever the protocol changes. Questions about safety are welcome through the support form. How we handle the data involved is in our privacy policy.

Effective September 24, 2026.