Kayts

Privacy Policy

This policy explains what personal data Kayts collects, why, how long we keep it and what rights you have. We wrote it to be read, not skimmed past. If anything is unclear, write to us (section 1).

1. Who is responsible for your data

Kayts ("we") is responsible for your personal data. You can reach us at privacy@kayts.ai about anything in this policy.

2. What we collect

DataWhere it comes from
Account: email address, and the name and profile details your sign-in provider shares (Google, Apple, Microsoft or ORCID)You, when you sign in
Profile: the name you give us and your date of birthYou, when you set up your account
Conversations: your messages, files and images you attach, papers and files you add in Research (their text is kept with the conversation), and the answers Kayts writesYou, when you use Kayts
Private chats: the same as conversations above, marked as private. A private chat doesn't appear in your chat history or search, can't be shared, and Kayts doesn't use or update Memory in it. It is still stored on our servers, like your other conversationsYou, when you start a private chat
Memory (Pro and Max plans): short facts about you that Kayts picks up from what you type, such as your job or how you like answers written. Never contact details, passwords or ID numbers. Health, religion or beliefs, political opinions, sex life or orientation, ethnic origin or trade union membership only if you are an adult and turn on “Include sensitive topics in memory” in Settings → Memory (off by default)Kayts, from your messages, while memory is on
Recordings you ask about (Audio): the sound of a recording you ask Kayts about, at most its first 10 minutes, converted in your browser to a small copy. The file itself stays in your browser. We don't store the sound or the answer; the answer is only shown to youYou, when you choose Describe, Transcribe or ask a question about a recording
Shared links: which conversations you shared, and up to which messageYou, when you share a chat
Usage: how many tokens (pieces of text, or of sound for Audio) you use per model in the current 24-hour and 7-day windows, and your planOur server, while you use Kayts
Payments: for a paid plan, your plan and the IDs of your customer record and subscription at Stripe. Your card or other payment details go straight to Stripe; we never see them. Link, which sells the paid plans, keeps the payment, billing address and invoice details it needs as the sellerYou, Stripe and Link, when you buy, change or cancel a plan
Sign-up protection: a hashed (unreadable) form of your IP address and email address from when you set up your account, so one person can't collect the free allowance over and over with new accountsOur server, when you set up your account
Consents and choices: which version of the Terms and this policy you accepted and when, and whether conversations may be used for trainingYou, when you set up your account and in Settings
Waitlist: your email address, the language of the page, when you joined and when you confirmed the address, and until you confirm it a hashed (unreadable) form of your IP addressYou, when you join the waitlist on our home page before Kayts opens
Chat without an account: the messages you send and the answers Kayts writes; a random identifier and how many free messages you have used, in a cookie in your browser (kayts_guest); and a hashed (unreadable) form of your IP address, to limit messages without an account from one networkYou and your browser, when you try the chat without signing in
Technical data: IP address, browser type and error logsYour browser, automatically
Visit statistics, only if you allow them: which page of kayts.ai you opened, the page you came from, your browser, type of device and country, and how fast the page loaded. Counted by Cloudflare Web Analytics without cookies and without an identifier that recognizes you on a later visit or on other sitesYour browser, after you choose "Allow statistics" in the cookie notice

Settings such as theme or text size, and the "About you" text in Personalization, are stored only in your browser. The "About you" text is sent with each message so answers fit you, and is not stored on our servers.

The kayts_guest cookie is needed for the chat without an account to work and to keep its limit, so we set it without asking for consent. It holds no account, name or email, and we don't use it to follow you anywhere else. You can delete it in your browser at any time; you then start over as a new visitor.

Your browser's storage also keeps your sign-in, your settings and your choice in the cookie notice. They are needed for Kayts to work, so we don't ask about them either. The one thing we ask about is visit statistics: the notice at the bottom of the page offers "Only necessary" and "Allow statistics", equally easy to choose, and nothing is counted until you allow it. You can change your choice anytime under "Cookie settings" at the bottom of every page. The desktop app counts no visits at all.

3. Why we use it, and on what legal basis

PurposeLegal basis (GDPR)
Creating your account, answering your messages and your questions about recordings, keeping your chat history, applying your plan's limitsPerforming our contract with you (art. 6(1)(b))
Memory: remembering facts about you to use in new chats, while you keep it onPerforming our contract with you (art. 6(1)(b)); you can switch it off and delete what it remembers anytime
Memory of sensitive topics: remembering the special categories of personal data listed above, only while you keep “Include sensitive topics in memory” onYour explicit consent (art. 9(2)(a) and art. 6(1)(a)); we record when you gave it. Turning it off withdraws your consent and deletes those facts
Showing a chat you shared to anyone who has its linkPerforming our contract with you (art. 6(1)(b)), at your request
Waitlist: emailing you a link to confirm your address, then writing to you once when Kayts opensYour consent (art. 6(1)(a)), given with the checkbox when you join and confirmed with the link. You can withdraw it anytime by writing to privacy@kayts.ai; we then delete your address
Waitlist: keeping out automated sign-ups (a check that tells people from scripts, and a limit on sign-ups from one IP address)Our legitimate interest in a list of real people (art. 6(1)(f))
Chat without an account: answering the few messages you send before you create an accountProviding the service you ask for (art. 6(1)(b))
Chat without an account: limiting how many messages one browser and one network can send, and keeping out automated use (the cookie, the hashed IP address and the Turnstile check)Our legitimate interest in a free trial that people, not scripts, use, at a cost we can carry (art. 6(1)(f))
Checking your age, and whether a parent's consent is neededLegal obligation (art. 6(1)(c) and art. 8)
Keeping the service secure, preventing abuse, fixing errorsOur legitimate interest in a safe, working service (art. 6(1)(f))
Visit statistics: how many people visit the website, which pages they read and which load slowlyYour consent (art. 6(1)(a)), given in the cookie notice. You can withdraw it anytime under "Cookie settings"
Keeping conversations for, and later fine-tuning, Kayts models on them, if "Help improve Kayts" is on (section 4)Our legitimate interest in improving the service (art. 6(1)(f)); for users under 18 only if they switch it on themselves
Selling you a paid plan: taking payment, renewing it each month, changing or cancelling itPerforming our contract with you (art. 6(1)(b))
Billing and accounting: invoices and tax records for paid plansLegal obligation (art. 6(1)(c))

4. Training on conversations

Kayts uses several language models (section 5). To make Kayts better at what people actually ask it, we plan to fine-tune open models, whose licenses allow it, on real conversations: further training that adjusts an existing model with examples, rather than building a new one from scratch. We are not training on conversations yet. Fine-tuning starts in a later stage of development; until then, conversations of accounts with "Help improve Kayts" on are kept so they can be used once it does. Before that happens we will update this section with the details of the first run.

Who is included

What is used

How it works

  1. No separate copy while you chat. Training data is taken from chat history only when a training run is prepared. If you delete a conversation or switch training off before then, it is simply not there.
  2. Selection. We take conversations of accounts that allow training at that moment and leave out everything listed under "Not used".
  3. Cleaning. We automatically remove identifiers we can detect, such as email addresses, phone numbers, national ID numbers (such as PESEL), card and bank account numbers, and IP addresses, and replace them with placeholders like [EMAIL]. Automatic cleaning can miss things, so please don't share secrets or other people's personal data in chats.
  4. Training. The cleaned data set is not linked to your account or name. It is copied to rented graphics processors (RunPod, section 5) only for the training run and deleted from them afterwards. The data set is never published or sold.
  5. Record. We record which conversations went into which training run, so we can honor your choices later.

The trained model answers all Kayts users. Conversations already used in a finished training run cannot be taken back out of that model, but once you switch training off or delete them they are never used again.

5. Who else processes your data

We do not sell your data and do not share it for advertising. We use these service providers, bound by data processing agreements:

Voice typing and voice mode use your browser's built-in speech recognition. In Chrome and Edge your voice goes to Google or Microsoft under their own privacy policies; Kayts only receives the text. In voice mode the answers are read aloud by ElevenLabs, or by your browser's own voice once your plan's allowance for it is used up.

Several providers are outside the European Union, most of them in the USA. Where data leaves the EU, transfers are protected by the EU-US Data Privacy Framework or the European Commission's Standard Contractual Clauses (art. 45 and 46 GDPR).

6. How long we keep it

7. Children

Kayts is not for children under 13. In Poland, the age of digital consent is 16 (art. 8 GDPR), so users aged 13 to 15 need a parent's or guardian's consent. We are still building the way to collect it; until then, accounts of users under 16 are deleted straight away and no data is kept. To use the chat without an account you must be at least 16. If you believe a child under 13 has given us data, write to us and we will delete it.

8. Your rights

You have the right to:

You can use some of these rights yourself in the chat: Settings → Account lets you correct your name, download a copy of all your data (a JSON file you can take elsewhere) and delete your account with everything in it; Settings → Memory shows what Kayts remembers about you and lets you delete it. For anything else, write to privacy@kayts.ai. We answer within one month.

9. AI and automated decisions

Kayts is an AI system and its answers are generated automatically. They can be wrong and are not professional advice. We do not make decisions about you with legal or similarly significant effects based solely on automated processing.

10. Security

Data is encrypted in transit (HTTPS) and at rest. Each account can access only its own data, enforced by the database itself (row-level security). Only a small number of people at Kayts can access the systems, and only when needed to run them.

11. Changes

If we change this policy in a way that matters, we will tell you in Kayts or by email before the change applies. The version and date are at the top of this page.