How We Use AI
Version 2026-09-27 · Effective 27 September 2026
In Viyu, AI works in two places only: the assistant and letter drafts. It's optional and switches on only if you give explicit consent. The Turkish version prevails if the two differ.
What for
- Answering your visa and travel questions and explaining rules in plain language.
- If you ask, drafting general texts such as a cover letter or an employer letter. Your name, passport number and signature come as empty tags in the draft; you fill them in on your phone.
Rules, dates, fees and sources don't come from AI; they come from Viyu's verified rule base. AI only explains. It doesn't read or check your documents; documents never leave your phone.
Provider
We use a third-party AI provider: DeepSeek. The provider is outside Türkiye; its servers are in the People's Republic of China. If the provider changes, we update this page and ask for your consent again.
What is sent
- The question you type, the latest messages of the same chat (at most 12) and the short note you add for a letter.
- Only these basics of your plan: destination country, visa type, travel date range by month, your stage and your experience level.
- The public rule-base records relevant to the question.
What is never sent
- Your documents, their photos and PDFs, and anything read from them.
- Your and your family members' names.
- Passport numbers, national ID numbers, IBANs, account and card numbers.
- Addresses, phone numbers, e-mail addresses.
- Income and balance amounts.
- The rest of your profile: province, birth year, work, social-security and income details, passport, visa and refusal history, family details, document types and dates.
Two gates
- Names are replaced with labels (such as [AD_1]) before your message leaves your phone.
- Before anything goes to the provider, our server scans it again for patterns such as ID numbers, passport numbers, IBANs, card numbers and MRZ lines. If it finds one, nothing is sent and you're asked to remove the number and ask again.
The provider may retain this data and use it to improve its service under its own terms. So please don't type ID, health or criminal record details into the assistant.
What we keep
- Our servers don't store your questions, the assistant's answers or letter drafts.
- A content-free usage record (feature, model, duration, result) is kept for 13 months per call.
- If you report an answer, that question and answer are kept for 90 days.
Limits
- AI answers can be wrong. Every answer comes labelled Certain, Likely or Uncertain, with its source. When the source isn't clear, the assistant says so.
- The assistant never predicts an outcome and never gives a score. Only the relevant authorities decide on visas.
- The assistant doesn't give legal advice and is not a lawyer.
- A change the assistant suggests is never written to your profile without your approval.
- Long-press an answer and choose "Report as wrong or inappropriate" to send it to us.
Your control
- The assistant works only with your explicit consent. Consent text: Consents.
- Without consent, your checklist, vault, Day Ledger, travel log and reminders keep working, on paid plans too. Features that use AI (the assistant and letter drafts) don't work; we say so on the plan screen before you buy.
- Withdraw your consent anytime in Profile → Privacy & data. Nothing new is sent to the provider after that.