AI safety

You are talking with an AI, not a person

The assistant is an AI. It is not a person, and it is not a doctor, nurse, therapist, counselor, dietitian or any other licensed professional. Nothing it writes is provided or endorsed by one. If you ask whether you are talking with a person, it will tell you no.

Every chat starts with a notice that you are chatting with an AI, and a long conversation repeats it every three hours. Under every chat, a line says that what the assistant writes is AI-generated, and the daily note on your home screen is labeled AI-generated too.

What it does

It gives general wellness education, using your own data so the answers are about you.

• Explains your results, wearable trends, sleep, meals, training and check-ins in plain words.
• Talks through habits, nutrition, exercise and sleep.
• Helps you log what you eat, take and do, and keep track of your goals.
• Helps you prepare questions to take to your own clinician.

What it does not do

It does not diagnose, treat or prescribe, and it does not tell you what dose of a medication to take. It is not therapy or counseling. It can be wrong.

It is not a substitute for clinical care. Talk with a licensed clinician about any condition, and before you start, stop or change a medication or a treatment.

What it remembers, and how to delete it

It keeps the lasting facts you share between conversations, like your goals, your history and how you describe what you feel, so you never have to start over. It writes each one as a note you could read, because you can.

You can read every memory, and delete any single one, on the What Valdura remembers about you page (Memory, on the web and in the app). You can also ask the assistant to forget something. You can remove a conversation from your chat list, and deleting your account deletes all of it.

Open what Valdura remembers about you

For adults 18 and older

Valdura is for adults 18 and older, and it may not be suitable for some minors. We ask for your date of birth before you start. If the date given is under 18, the account is closed: no chat and no app. We do not keep that date.

The crisis protocol

The assistant watches every conversation for signs that someone may be in danger, and puts that ahead of everything else it was asked to do.

• What it watches for: thoughts of suicide; self-harm, including eating disorders such as purging or starving yourself; thoughts of harming someone else; a medical emergency; and a possible overdose or poisoning.
• What it does: it names the right help in its first sentence, before anything else. It offers 988 by call or text, with the link to 988lifeline.org, and the Crisis Text Line (text HOME to 741741). It says to call 911 for immediate danger, and Poison Control at 1-800-222-1222 for a possible overdose or poisoning. It takes what you say seriously the first time, asks whether you are safe right now, stays with you, and does not turn the conversation to data or products.
• What it never does: Valdura never contacts the police, or anyone else, about you. It does not send a crisis conversation to law enforcement. It points you to people trained to help, and the choice to reach out is yours.
• Always on screen: the crisis lines sit under every chat, on every signed-in page of the website, and in the app's You tab.
• In the community: a post or a question that signals a crisis is not posted. You see the crisis lines instead.

How we count crisis referrals

Some states ask companies like ours to report each year how many times they referred someone to a crisis line. So Valdura counts each time it shows 988 or the Crisis Text Line, 911 or Poison Control: a number per day, by kind and by where in the product it happened. No name, no account and no words are kept with the count.

What powers it

Anthropic's Claude, running in the United States under a Business Associate Agreement, on models that do not train on your data. Anthropic holds what it processes briefly, up to 30 days, then deletes it.

Privacy Policy