Safety

Last updated: October 23, 2026

Version: us-2026-10b

If you're in danger right now

FP AI can't help in an emergency, and it can't send help. Talk to a real person now:

Immediate danger: call 911.
Thinking about suicide or hurting yourself, or in emotional crisis: in the US and Canada, call or text 988, free and 24/7. In the US you can also chat at 988lifeline.org.
Someone is hurting, threatening or controlling you: National Domestic Violence Hotline, 1-800-799-7233, 24/7, or thehotline.org.
UK and Ireland: Samaritans, 116 123.
Australia: Lifeline, 13 11 14.
Anywhere else: call your local emergency number.

FP AI is an AI, not a person

FP AI is an AI coach built on Fred's method. It is not Fred, not a human coach and not a professional, and no person writes its replies.
When you start chatting, and at least every 3 hours while you keep chatting, the app shows this notice: "You're chatting with FP AI, an AI coach built on Fred's method. It's not a person."
If you ask FP AI whether it's an AI, a real person or Fred, it tells you the truth.

What FP AI is not

Not an emergency service or a crisis line. It can't call anyone, send help or contact emergency services for you.
Not medical care, mental health care or counseling.
Nobody watches your conversations in real time.

How our safety protocol works

We built FP AI to stop coaching and send you to real people when you may be in danger. This is what actually happens.

1. Word check. Every message you type, including in temporary chats, is checked against a list of words and phrases that can signal suicide, self-harm, abuse or violence against you, or that you may be under 18. Voice notes are checked the same way once they're turned into text.

2. Safety check. If something matches, a separate AI safety check reads the words around it, to tell real danger, or a sign that you may be under 18, from figures of speech like "this is killing me," "I'd rather die than text him first" or "I'm 15 minutes away." If the safety check can't run or doesn't answer in time, we treat the message as a crisis: when in doubt, we show the resources.

3. Safety message. If the risk is confirmed, FP AI doesn't coach on that message. Instead of a normal reply, it shows a fixed message written by us, not generated by AI, with the numbers at the top of this page.

4. Under 18. If the safety check finds that you may be under 18, FP AI stops coaching, shows a fixed message for minors and puts the account on hold until a person on our team reviews it (see "Adults only" below).

5. Limits and payments don't get in the way. The safety message is shown even if you've reached your daily limit, and it never waits for a payment or consent check.

6. Our team may be alerted. The app may send our team an internal alert that the safety message was shown on your account, or that the account was put on hold. The alert doesn't include what you wrote. It isn't an emergency service and doesn't send help to you. [TO BE CONFIRMED: who reviews these alerts and whether we may contact you by email.]

7. No selling. For 24 hours after a safety message, FP AI shows no offers or promotions of any kind.

8. Beyond the word list. FP AI's own instructions tell it to do the same whenever it recognizes danger the word check can't catch, for example in a screenshot or in words that aren't on the list: step out of coaching, say clearly that this goes beyond advice about a man, and share the same numbers. It's instructed never to make up or change a number.

What the protocol can't do

No system catches everything. The word check reads typed text and transcribed voice notes, not images. For screenshots, FP AI's own judgment is the safety net.
If you've reached your daily limit, voice notes can't be processed. Type your message instead, or call or text 988.
FP AI doesn't know where you are and can't send anyone to help you. If you're in danger, call 911.
FP AI answers in English, and the word list is mainly in English. If you write in another language, the word check may miss it, and FP AI's own instructions are the safety net.

Adults only

FP AI is for adults 18 and over. Everyone confirms their age before their first conversation.
If our checks find that you may be under 18, for example because you say so, FP AI stops coaching. Instead of a reply, it shows a fixed message written by us, which explains that FP AI is only for adults, suggests talking to an adult you trust and gives the 988 number.
The account is then on hold: FP AI doesn't answer messages or voice notes until a person on our team reviews it. If you're in danger, call or text 988, or call 911.
If you're 18 or older and your account was put on hold by mistake, email supporto@federicopicchianti.com and we'll review it.
If we confirm that an account belongs to someone under 18, we close it and delete its data.
Companion chatbots may not be suitable for some minors.

Your privacy when the safety message is shown

We keep a record that the safety check ran, its result, when the safety message was shown and, if it happens, when the account was put on hold.
The message that triggered it and the safety message stay in your chat history, unless you were in a temporary chat. You can delete them like any other conversation.
If an internal alert is sent, it identifies your account, usually by email, and never includes your messages.
Our anonymous daily counts of safety events (see below) aren't linked to your account.
We don't share your conversations with anyone because of a safety message, unless the law requires it or it's needed to protect someone's life.

More details in our Privacy Policy and in our Consumer Health Data Privacy Policy.

Measuring and improving the protocol

We count how often the safety check runs, how often it confirms a risk and how often the safety message is shown, without reading your messages.
For the yearly report, we also keep a daily tally of safety events, such as how many times the safety message was shown. It has no names, email addresses, account numbers or message content, so it stays even after an account is deleted.
We use these numbers to improve the protocol, and we report them where the law requires it (for example in California's annual report to the Office of Suicide Prevention), without names or message content.
We review the word list, the safety check and the numbers on this page regularly. [TO BE COMPLETED: review frequency]

Questions or concerns

If something FP AI said worried you, or you want to report a safety concern, email supporto@federicopicchianti.com. This inbox isn't monitored around the clock and can't respond to emergencies. If you're in danger, use the numbers at the top of this page.