KULT AI AGENCY · KULT GENERATE
Privacy Policy
Version of 29 October 2026
This policy explains which personal data we process when you use the KULT AI AGENCY website (kultai.agency/production/) and the AI studio KULT GENERATE (kultai.agency) with its Telegram bot @kultaibot, its apps for iPhone and Android and its API, why we do it, who else receives it and what rights you have under the EU General Data Protection Regulation (GDPR). Section 6 is our cookie policy. We have tried to keep it plain.
1 Who is responsible
The controller for the processing described here is KULT AI AGENCY.
Operator: Karen Grigoryan, a private individual resident in Spain, trading under the name KULT AI AGENCY. NIE: Z4712786E. Address: C/ Jesús 69, 2nd floor, door 4, 46007 Valencia, Spain. E-mail: support@kultai.agency.
For any privacy question or request, contact us:
- on Telegram: @kultaibot — the fastest way;
- by e-mail: support@kultai.agency;
- by phone: +34 624 02 95 85.
2 What we process and why
2.1 Agency website and order builder
When you send the contact form or the order builder, we receive your name, the way you would like us to reach you (e-mail, Telegram or WhatsApp) together with the address or username for it, your answers about your company, industry and project, and the messages you write. We store the request on our server and forward it to our team via Telegram, so that we can reply and prepare an offer.
Until you send a form, what you have typed stays in your browser (see section 6).
2.2 Your KULT GENERATE account
You can register with your e-mail address and a password, or sign in with Google or Telegram and, where the studio shows them, with Apple or Microsoft. With e-mail registration we store your e-mail address; your password is stored only as a salted hash, never in readable form. When you sign in through a provider, we receive your account ID with that provider and, if the provider shares them, your name and e-mail address.
In your profile we also store your display name, your interface language and the colour of your avatar.
Sign-in with Google. We use the data Google shares with us (account ID, name, e-mail address) only to create and run your account, to sign you in and to prevent the welcome bonus from being claimed more than once (see section 7). We do not use it for advertising and we do not sell it. The studio’s Google button takes you to Google’s own sign-in page and then back to the studio. Only where that is not available is Google’s sign-in script (Google Identity Services) loaded from Google’s servers, and only when you open the sign-in window.
Sign-in with Apple. Apple sends us an ID of your Apple account that is valid only for our service, your e-mail address and, the first time only, your name. If you choose Hide My Email, Apple gives us a private relay address (…@privaterelay.appleid.com) that forwards to your real address; we use it as your account’s e-mail address. We keep the name for up to 24 hours until your account is created, so that it is not lost if the sign-up is interrupted.
Sign-in with Microsoft. Microsoft sends us the ID of your directory (tenant) and your object ID in it, which together identify your account, and your name and e-mail address. For a work or school account we use the e-mail address only if Microsoft confirms that your organisation owns its domain, and such accounts are not treated as verified. We use the data from Apple and Microsoft only for the same purposes as the data from Google.
Sign-in with Telegram. Telegram sends us your Telegram ID, your name, your username and, if you have one, a link to your profile photo. We store the username with your account and use the photo as your avatar. Telegram’s sign-in widget is loaded from telegram.org only when you choose to sign in with Telegram. When you use our bot or its Mini App (2.7), Telegram passes us the same data.
Profile photos. If your avatar is your Google or Telegram profile photo, your browser loads it directly from Google’s or Telegram’s servers, which see your IP address when it does. From Google we also store the link to the photo and, if your Google account belongs to an organisation (Google Workspace), the domain of that organisation.
2.3 KULTCOINS, payments and purchase requests
We keep your KULTCOINS balance and its history (credits, spending, refunds and expiries), and for each credit the date its KULTCOINS expire (Terms, section 4.6), because this is how you pay for generations.
When you pay for a plan, a KULTCOINS pack or a single generation, the payment is made on a page of our payment provider Stripe. Stripe collects your card or other payment details and your billing address, which it keeps in your customer record at Stripe; we never see or store your card number, and we do not store your billing address on our server. From Stripe we receive and keep what we need to credit your purchase and for accounting: the Stripe customer and payment IDs, the item, the amount and currency, the status of the payment and of your subscription (renewal, cancellation, refunds and disputes), your consent to immediate delivery and the fingerprint of the card used (a code that Stripe derives from the card number: it recognises the same card again but does not reveal the number), which we use against self-referrals and abuse (2.6). When we create your Stripe customer record we send Stripe your account ID and, if you have them, your name and e-mail address. For a single generation we keep the generation’s settings until it has run.
If online payment is unavailable and you send a purchase request instead, we record the request: the plan or pack, its price and any promo code.
Cancelling a subscription. If you cancel through the cancellation form (“Cancel subscription”, no sign-in needed), we record the e-mail address and name you enter, the kind of cancellation (ordinary, or for good cause), your note, the date and time we received it and its confirmation number. We look up the account with that e-mail address and its subscription; if exactly one subscription matches, we stop its renewal at Stripe at once. The request, with the matching account, is forwarded to our team via Telegram (section 4). If you cancel in your signed-in account, we record the confirmation number, the subscription, its end date and the time. We use these records to carry out your cancellation and as proof of it (how long: section 7).
2.4 Your content
To run generations and chats we process the reference files you upload (images, video, audio including voice recordings, and PDF documents in the chat), your prompts and the texts you have read out by a voice, the generated results, your conversations with AI models in the chat, and the way you organise your work (favourites and folders). Sections 2.9 to 2.11 describe what the voice and video tools, characters, Marketing and KULT SUPERCOMPUTER add. If you press “Improve” next to a song’s style or lyrics, that text is sent through OpenRouter to an AI language model, which rewrites it (section 4).
2.5 Technical and security data
To operate the service and fix errors, we keep technical logs with metadata about requests. Our web server’s access logs record the IP address, the time and the address of each request and the browser type; our application itself does not store visitors’ IP addresses — only a keyed hash for the free included generations’ per-address limit, if we offer them (2.15). Technical logs are deleted after 30 days (section 7).
For each signed-in session we store when it started and was last used and the browser type (user agent), so that you can see your sessions in your profile and end them. A session ends after 90 days at the latest.
To protect the service we use rate limits and measures against abuse and fraud, for example to stop the welcome bonus from being claimed repeatedly.
We show no advertising, use no analytics trackers and never sell your data.
2.6 Referral & Ambassador Program
Every account has a referral code. For the programme we process your code and links and how many times they were opened; which accounts you invited and, if you were invited, who invited you and how (link, Telegram start link or code); the rewards and bonuses with their status; and the notices about them, which we show on the Partnership screen and, if your account is linked to Telegram, also send to your chat with @kultaibot.
If you apply to become an ambassador, we store the links to your social profiles, your audience size, your languages and your note, and forward them to our team via Telegram. If you claim a views bonus, we store the link to the video, the number of views you give and the screenshot of its statistics.
To keep the programme fair, we check invitations for self-referrals and fraud: we compare the accounts involved by the browser they signed in from (the known-device cookie, section 6), their Telegram account and the payment card used (its fingerprint, 2.3). Matches are recorded for our review and can lead to a reward being withheld (Terms of Use, section 15).
2.7 Telegram bot and Mini App
You can also use KULT GENERATE in Telegram, through our bot @kultaibot and its Mini App. Your prompts, the photos you send and the results then pass through Telegram, which processes them under its own privacy policy. For this we store the draft of the request you are preparing in the bot (prompt, settings and attached files), the buttons under recent messages and which generations were started from which chat.
2.8 Support
If you write to us in the support mode of @kultaibot, your messages are forwarded to the Telegram accounts of our support staff together with your Telegram name, username, ID and language and, if your Telegram is linked to a studio account, that account’s number, e-mail address, plan and balance, so that we can help you quickly. For the support chat we keep a contact entry with your Telegram ID, name, username and language (for how long: section 7). To deliver our replies to your chat, we keep the link between a forwarded message and your chat for 60 days. E-mails to support@kultai.agency reach our team’s mailbox through the e-mail forwarding service of our domain provider (section 4).
Support assistant. “Help” in the studio and in your profile opens the KULT support assistant. It answers questions about the studio, prices, plans, KULTCOINS and these policies with the matching help article or, for a question about a price, with the current prices of our models. If you are signed in and ask about a generation of yours that failed or got stuck, or about KULTCOINS you are missing, it shows the status of your last three generations (model, status, KULTCOINS charged and returned). These answers are made without a language model, and your question is not sent to any AI provider. When AI answers are switched on, the assistant can also answer signed-in accounts with a text written by an AI, and the chat window marks each such answer. To write it, our server sends through OpenRouter to the language model Grok of xAI (USA) (section 4) your question, the earlier messages of the same conversation, the matching help articles, the studio’s current data (prices, which models are available, known disruptions) and, if your question is about your plan, KULTCOINS or generations, your plan (with its end date if you ask about it), your KULTCOINS balance and your recent generations (up to five, and the one you ask about: model, how long ago, status, kind of error, KULTCOINS charged and whether they were returned; never their prompts or results). Before sending, it removes e-mail addresses, phone numbers and payment details (such as card or account numbers) from the text; your account’s e-mail address, name and payment data are never sent. The assistant cannot refund money, credit KULTCOINS, give promo codes or change your account. If it cannot answer, it offers to pass your question to a person; a support request with your question, your account and the conversation is created when you use “Talk to a person”. If you are signed in and ask for a refund of a payment, want to withdraw from a contract or make a complaint, the assistant passes your message to our team by itself, as such a support request; when AI answers are switched on, it does the same with a dispute about money for a particular payment. Our team is told via Telegram without your e-mail address (section 4) and replies to the e-mail address of your account, usually within one business day. In the Telegram Mini App, or if you sign in with Telegram and your account has no e-mail address, “Talk to a person” opens the support chat of our bot instead. Under each answer you can mark whether it helped. To improve our help articles we count, without linking them to you, the topics of questions, the questions the assistant could not answer and the ratings of its answers. How long we keep this: section 7.
2.9 Voiceover, talking videos and background removal
Remove background. The picture you choose is sent through kie.ai to Recraft, which cuts out the subject; our server saves the result as a PNG file.
Voiceover. The text you want read out (or the lines of a dialogue), the voice you pick from our catalogue and its settings are sent through kie.ai to ElevenLabs, which returns the audio. You can only choose voices from the catalogue: we do not copy or clone anyone’s voice. We keep the text with the generation, like a prompt.
Talking video. For a talking video you upload a photo of a person (for one of the models, a video), add a voice (a recording you upload or a voiceover made in the studio) and, if you like, describe the movement. These files and texts are sent through kie.ai to the developer of the model you choose: Kling Avatar (Kuaishou), InfiniteTalk (MeiGen-AI), OmniHuman (ByteDance) or the lip-sync model of Volcengine (ByteDance’s cloud service). Our server may convert an uploaded voice recording to MP3, without its metadata, and may make a trimmed or resized copy for the provider. Before you upload a face or a voice, the studio asks you to confirm that you hold the rights and that the people concerned agree (section 9).
2.10 Characters (Influencer) and Marketing
Characters. In the Influencer builder you create a character either from a description or from a photo of a person. From a description, you choose features such as gender, age, origin, skin tone and style; they describe an invented person, and the default “Mix” sends the description to four image models at once (Higgsfield’s Soul and models of Google, OpenAI and ByteDance). From a photo, the studio first asks you to confirm that it is your photo or that you have the consent of the person shown, who must be 18 or older; it then does not ask for or send features such as origin or skin tone, and the default “Mix” sends the photo through kie.ai or Higgsfield to four models at once: Nano Banana Pro (Google), GPT Image 2 (OpenAI), Seedream (ByteDance) and Flux (Black Forest Labs). You can choose a single model instead. For each character you save, we store in your library its name, how it was made, the features you chose, its images and, for a character made from a photo, your confirmation of consent with its time and the version of the Terms of Use.
Marketing. For a Marketing template you upload product photos, a logo and, if you wish, a photo of a person (with the same confirmation of consent), and you describe your offer. We process them like other uploads and prompts (2.4) and send them to the models of the template you choose. Until you start the template, your brief stays in your browser (section 6).
2.11 KULT SUPERCOMPUTER
KULT SUPERCOMPUTER turns a task you describe into a plan of generations. To make the plan, we send through OpenRouter to the provider of the planning model (for example OpenAI, Anthropic, Google or xAI in the USA, or DeepSeek in China): your task, the last messages of the conversation, the active brand profile, if there is one, and a short description of your files (type, size, duration), never their names or contents. The generations of the plan then run like any other generation (2.4, section 4).
- Brand profiles: up to five profiles with a brand’s name, description, audience, tone, colours, logo, rules and preferred format and language. An active profile is sent with each planning request.
- Web search stays off unless you switch it on. Then OpenRouter sends a search query made from your task to its search provider Exa (section 4), and we store the sources found with the answer.
- “Run right away”: if you choose this mode, the planned generations start without a further tap, within the limits you set. We store the mode, your limits and which runs started automatically.
- We store your plans, their steps and the files linked to them with the conversation, and the amount of each run in your KULTCOINS history.
2.12 API access
If you use our API (Terms of Use, section 18), we also process: your API keys, which we store only as a hash, with their name, last four characters, spending limits and the times of creation, last use and revocation (the full key is shown to you once and is never stored in your browser); the webhook address you set and the secret used to sign notifications to it; for each request, the key used and the callback address and metadata you send with it (the request itself is stored like any generation); for 24 hours, the answers to requests sent with an Idempotency-Key, so that a repeated request is not charged twice; the delivery records of webhook notifications; and a log of changes to your keys and webhooks, of your acceptance of the API terms and of changes our team makes to your API access. Your IP address is used only briefly in memory to enforce request limits; our application does not store it. When an account creates its first API key, our team is told via Telegram. If you send us personal data of your own customers, section 14 applies.
2.13 Checks of requests and reports
Automatic check of requests. Before a generation is started or a chat message is sent, our server checks the text automatically against a list of words and phrases that point to sexual content involving minors or to undressing a real person from a photo; in the apps also to explicit sexual content (section 13). A matching request is refused before any KULTCOINS are taken. We do not store the text of a refused request. We count refusals per account only in memory, for 24 hours; after three refusals within 24 hours our team is told via Telegram (the account, without the texts), so that a person can look at the account. The AI providers also check prompts, uploads and results with their own moderation.
Reports. With “Report” on a result, a generation, a chat answer or a preset, you can tell us what is wrong. We store your account, what you reported, the reason you chose, your note (if you write one), and the status and outcome of our review. Reports about minors or other illegal content reach our team via Telegram at once, including your note; the others are counted in a daily summary. How we handle reports and notices is described in section 17 of the Terms of Use.
2.14 E-mails about your contract and withdrawals
Service e-mails. We send you only the e-mails your contract and the law require: the confirmation of a purchase (with the details of the order, your consent to immediate delivery and its time, the withdrawal information and our Terms of Use as an attached file), the acknowledgement of a withdrawal, the confirmation of a cancellation and, if you ask for it, a link to reset your password. We send no advertising. For this we process your e-mail address (the one you gave Stripe for the payment, otherwise the one of your account), your name, your language and the content of the e-mail. Each e-mail stays in our outbox until it has been sent, with its recipient, subject, status, the number of attempts and any error; after sending we keep this record as proof that the confirmation was sent (section 7). The e-mails are sent through our e-mail provider (section 4).
Withdrawal from a contract. If you use “Withdraw from contract” (Terms of Use, section 5.6), we record your name, the e-mail address you enter, the contract you name (or the purchase you pick when signed in), your message, your language, the date and time we received it and its reference number, and, if you are signed in, your account. We send the acknowledgement to the address you enter — with the details you typed only to an address of your account that we know — and check whether the address belongs to a customer; our team is told via Telegram, with a suggested outcome and links to the payment at Stripe. If you withdraw from a plan while signed in, we stop its renewal at Stripe. A hidden field of the form catches automated spam; such submissions are stored separately and deleted after 30 days, unless the e-mail address or the signed-in account is a customer’s — then the declaration is handled as usual. We use these records to handle your withdrawal and as proof of it (how long: section 7).
Password reset by e-mail. If you ask for a reset link in the sign-in window, we send it to the e-mail address of the account, if such an account exists; the answer on screen is the same in either case. The link works once and for 1 hour; we store only a hash of it, and the link is removed from our outbox as soon as the e-mail has been sent.
2.15 Prices, personal offers and included generations
Personal offers. To decide whether your account is shown a personal offer for a yearly plan (Terms of Use, 5.1), we use fixed rules on data we already hold: whether the account has ever bought a plan, whether it holds one now and since when it has none, and the offers it has had before (their kind, code, deadline, whether used). Everyone in the same situation gets the same price; the price is not calculated for you personally. The offer is shown only inside the studio — we send no e-mail, push or bot message about it. We keep the offer records while your account exists, so that an offer is used once. Legal basis: our legitimate interest in offering our plans to people who already use the studio (Art. 6(1)(f) GDPR); you can object at any time (section 8), and we then show you no personal offers. This is no decision under Art. 22 GDPR: the regular prices stay open to you either way.
Included generations, first month and trial. To count the generations included in a plan (“up to N a day”), we record for each such generation the account, the model and the time, for 60 days. Only if we offer included generations to accounts without a plan do we also store a keyed hash of the IP address (never the address itself), to apply a per-address limit, for 60 days. For the first-month price and the trial we check your past plan payments and, for the trial, the fingerprint of the card that Stripe gives us (not the card number), so that each is used once per account and per card. Legal basis: the contract (Art. 6(1)(b) GDPR) and, for the abuse checks, our legitimate interest (Art. 6(1)(f) GDPR).
3 Legal bases
- Contract and pre-contractual steps (Art. 6(1)(b) GDPR): your account, generations and chats, the voice and video tools, characters and Marketing, KULT SUPERCOMPUTER (including web search and “Run right away” when you switch them on), API access, KULTCOINS, purchases on the website and in the apps, purchase requests, the referral programme, the Telegram bot and support, including the support assistant (2.8), and replying to the request you send through the website. This includes sending your prompts and files to the AI providers in section 4, without which we cannot create what you ask for.
- Legitimate interests (Art. 6(1)(f) GDPR): security of the service, prevention of abuse and fraud (including the self-referral checks of the referral programme, 2.6, and the automatic check of requests, 2.13), the log of changes to API keys and webhooks (2.12), notices to our team, backups, personal offers inside the studio (2.15) and statistics of questions to the support assistant, not linked to you, to improve our help articles (2.8). Our interest is to keep the service, its users and your data safe and available. You can object to this (see section 8).
- Legal obligation (Art. 6(1)(c) GDPR): keeping accounting and tax records of payments, and handling notices about illegal content and giving reasons for our decisions under the EU Digital Services Act (2.13); sending and keeping the confirmations consumer law requires — of a purchase, of a withdrawal and of a cancellation (2.3, 2.14).
- Consent (Art. 6(1)(a) GDPR): only where we explicitly ask for it: for cookies and similar technologies that are not strictly necessary, which you allow in our cookie banner (section 6), and for push notifications (section 13). You can withdraw consent at any time with effect for the future, as easily as you gave it.
We do not process special categories of personal data, such as biometric data for identifying a person, for purposes of our own (section 9).
You are not obliged to give us any data. Without an e-mail address or a sign-in account, however, we cannot create a studio account, and without a way to contact you we cannot answer your request.
4 Who receives your data
We pass on data only where this is needed to provide the service. For each recipient we name its country, where we know it, and, for a transfer outside the EU/EEA, what the transfer rests on (details in section 5):
- Hetzner Online GmbH (Germany) hosts our server in a data centre in the EU; our databases and files are stored there. No transfer outside the EU.
- kie.ai (NEXUSAI SERVICES LLC and INNOLEAP AI LLC, Colorado, USA) runs most generations, the voice and video tools (2.9) and every request made through our API (section 14). It receives the prompt, the settings and the reference files of the generation, which often show faces and can contain voices, and sends back the result. Under its terms of use, kie.ai receives “a worldwide, non-exclusive, royalty-free license to use, host, store, reproduce, modify, and display” uploaded content, for operating and improving its services. Its privacy policy says that data may be stored in the USA. We have no data processing agreement with kie.ai. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- Higgsfield AI (Higgsfield Inc., San Francisco, USA) runs the other generations, for example with the Soul, Cinema Studio, Kling, LTX, PixVerse, Recraft, Ideogram, Qwen Image and Grok models, and receives the same data. Higgsfield’s API terms include its data processing agreement, and they allow Higgsfield to use content sent through its API to train and improve its AI models unless the API customer opts out. Transfer: under Higgsfield’s data processing agreement (section 5).
- BytePlus (BytePlus Pte. Ltd., Singapore, a company of the ByteDance group) runs Seedance video generations directly through its ModelArk service and processes them in Southeast Asia. It receives the prompt, the settings and the reference files of the generation, and a code derived from your account (not your name or e-mail address) that BytePlus uses to detect abuse. If BytePlus refuses a generation because a picture shows a real person, the studio offers it through kie.ai instead (at the price shown before you start). We have no data processing agreement with BytePlus. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- Developers of the AI models. Depending on the model you choose, kie.ai or Higgsfield may pass your prompt and files on to the developer of that model or to its cloud provider, which processes them under its own terms. These are mainly technology companies in the USA (for example Google, OpenAI, xAI, Suno and Topaz Labs) and in China and other Asian countries (for example ByteDance for Seedance and Seedream, Kuaishou for Kling, Alibaba for Wan and Qwen, MiniMax for Hailuo), and also in the EU (Black Forest Labs for Flux) and Israel (Lightricks for LTX). For the voice and video tools (2.9) they include ElevenLabs (Eleven Labs Inc., USA) for voiceovers, Recraft for background removal, and Kuaishou (Kling Avatar), MeiGen-AI (InfiniteTalk) and ByteDance (OmniHuman and its cloud service Volcengine) for talking videos. When you make a character from a photo, the photo goes to Google, OpenAI, ByteDance and Black Forest Labs at once, unless you choose a single model (2.10). We have no contract with these developers. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- OpenRouter (OpenRouter, Inc., New York, USA) powers the chat, the planning of KULT SUPERCOMPUTER (2.11), the “Improve” button for a song’s style and lyrics (2.4) and the AI answers of the support assistant when they are switched on (2.8). It receives your chat conversation, including files you attach there, the planning request, or the style or lyrics you ask “Improve” to rewrite, or, for an AI answer of the support assistant, your question with the earlier messages of that conversation (without e-mail addresses, phone numbers and payment details), the matching help articles, the studio’s current data and, if your question is about your plan, KULTCOINS or generations, your plan, your KULTCOINS balance and your recent generations (2.8), and routes it to the provider of the model (for example Anthropic, OpenAI, Google or xAI in the USA, or DeepSeek, Moonshot or Alibaba/Qwen in China; for the support assistant, xAI in the USA). According to its privacy policy, OpenRouter itself does not use your inputs and outputs to train models; the model providers process them under their own terms. We ask OpenRouter to route our requests only to model providers that do not store them or use them for training. OpenRouter’s terms make its data processing agreement part of our contract with it; we have no contract with the model providers. Transfer: under the Standard Contractual Clauses of that agreement; to the model providers, without an adequacy decision or Standard Contractual Clauses (section 5).
- Exa (Exa Labs Inc., San Francisco, USA) runs the web search of KULT SUPERCOMPUTER, only when you switch it on (2.11). Through OpenRouter it receives a search query made from your task. According to its privacy policy, Exa uses query data to improve its products, including to train models. We have no contract with Exa. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- Stripe (Stripe Payments Europe, Ltd., Ireland) processes payments, subscriptions and refunds. It receives the payment details you enter on its page and the data named in 2.3, and acts as an independent controller for its own legal duties (for example fraud prevention and anti-money-laundering checks). Stripe may transfer data to Stripe, Inc. in the USA; transfer: the EU–US Data Privacy Framework and Standard Contractual Clauses, as described in Stripe’s privacy policy.
- Google, Apple and Microsoft handle sign-in, only if you choose that sign-in method, as independent controllers. For users in the EU their Irish companies are responsible; data can reach their parent companies in the USA. Transfer: the EU–US Data Privacy Framework or Standard Contractual Clauses, as described in their privacy policies.
- For the apps (section 13): Apple and Google sell in-app purchases as the stores; RevenueCat (RevenueCat, Inc., USA) confirms these purchases for us as our processor; and if you switch on push notifications, Apple, Google (Firebase Cloud Messaging) or the push service of your browser delivers them. Transfers: section 5.
- Telegram (Telegram Messenger Inc. and its group companies, based outside the EU) handles sign-in with Telegram if you choose it, carries the use of our bot and its Mini App (2.7) and our notices to your chat (2.6), relays support chats to our support staff (2.8) and delivers notifications to our team, for example new requests from the website form, new payments, subscription cancellations (2.3), withdrawals from contracts (2.14), requests the support assistant passes to a person, without your e-mail address (2.8), ambassador applications, views-bonus claims, urgent reports including your note, repeated refusals by the request check (2.13) and new API users (2.12). Telegram processes these data under its own privacy policy. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- Spaceship (Spaceship, Inc., USA), our domain provider, forwards e-mails sent to our addresses, such as support@kultai.agency, to our team’s mailbox at an external e-mail provider, which stores them, possibly outside the EU. Transfer: without an adequacy decision or Standard Contractual Clauses (section 5).
- Resend (Plus Five Five, Inc., San Francisco, USA) sends our service e-mails (2.14) from its EU region (Ireland). It receives the recipient’s e-mail address and the content of the e-mail, including attachments, and keeps delivery logs under its own terms. Transfer: the Standard Contractual Clauses of Resend’s data processing agreement, which is part of its terms (section 5).
- Google (Google Drive) (Google Ireland Limited / Google LLC, USA) stores the nightly backup copies of our databases and media files, including uploads and results. Transfer: the EU–US Data Privacy Framework, under which Google LLC is certified.
When an API customer has its results delivered to its own servers, it chooses those addresses itself; they are not our recipients (section 14).
Each of these companies processes the data under its own terms and privacy policy. We do not sell personal data and do not share it with advertisers.
5 Transfers outside the EU/EEA
Our server is in the EU (Germany). Some recipients named in section 4, however, are based or process data outside the EU/EEA, mainly in the USA and, for some AI models, in China or other countries. When you use a function that relies on them, your data is transferred there:
- Stripe, Google (including Firebase Cloud Messaging), Apple and Microsoft transfer data to the USA, according to their privacy policies, under the EU–US Data Privacy Framework (adequacy decision of the European Commission of 10 July 2023) or Standard Contractual Clauses.
- Higgsfield: we rely on the safeguards of the data processing agreement that is part of Higgsfield’s API terms.
- OpenRouter, RevenueCat and Resend: we rely on the Standard Contractual Clauses (controller to processor) of their data processing agreements, which are part of their terms.
- kie.ai, BytePlus, the developers of the AI models (including the model providers to which OpenRouter routes a chat, a planning, an “Improve” request or a question to the support assistant), Exa, Telegram, Spaceship, our e-mail provider and the push service of your browser where it is not run by Google, Apple or Microsoft: there is no adequacy decision for these transfers, and no Standard Contractual Clauses have been agreed with us. We transfer data to them only as far as this is necessary to provide what you ask us for, for example to carry out a generation, a chat answer, a plan, an “Improve” rewrite, an AI answer of the support assistant or a web search, to deliver a push notification you switched on, to run our Telegram bot, to handle your request from the website or to receive your e-mail (Art. 49(1)(b) GDPR). The same applies to Higgsfield where its data processing agreement does not cover a transfer.
Countries outside the EU may not offer the same level of data protection as the EU: authorities there may have wider access to data, and your rights may be harder to enforce. If you do not want your data to be transferred in this way, do not use the function concerned; for example, do not upload photos or voice recordings of people.
7 How long we keep data
- Account: until you delete your account (in the studio or the app: Profile → Delete account) or ask us to delete it. When the account is deleted, your uploads, results, chats, characters, brand profiles, KULT SUPERCOMPUTER plans, API keys and webhook settings, push tokens, sessions and drafts in our Telegram bot are deleted, and so are the texts of your voiceovers and the callback addresses and metadata of your API requests. What we keep after that is listed below.
- Sessions: until you sign out, at most 90 days.
- Results: until you delete them or your account. We may set a storage period for results; while one is set, the results of an account without an active plan are kept at least 30 days after they were made, and the results of an account with an active plan (a subscription on the website or in the apps, or a plan we granted) are kept while the plan is active and at least 90 days after it ends. Never deleted this way: Favourites, results in your folders, the pictures of your characters, the files of a running KULT SUPERCOMPUTER plan, the logos of your brand profiles, files that an unfinished generation or API request uses, and your uploads. At least 7 days before results are deleted, we tell you in the studio and, if you use them, through our Telegram bot and by push notification; a result you add to Favourites or to a folder stays. If we change these periods, we update this section first.
- Deleted results: a result you delete from your library is removed from our server within 30 days.
- Uploads of new accounts: reference files uploaded by an account that is neither verified (Terms of Use, section 3) nor has made a purchase, and that were not used in any generation or chat, are deleted after 7 days.
- Photos, videos and voice recordings for talking videos, characters and Marketing are uploads like any other: we keep them until you delete them or your account; if a generation that uses the file is still running, we delete it when the generation ends. Copies we make for a provider (converted, trimmed or resized) are deleted together with the original. The text of a voiceover is kept with its generation, like a prompt.
- Characters and brand profiles: until you delete them or your account. When you delete a character, you can also delete its images. The confirmation of consent for a character made from a photo is kept as long as the character.
- KULT SUPERCOMPUTER: plans, their steps and the sources of a web search are deleted together with the conversation they belong to; the amounts of each run are kept like KULTCOINS records.
- API: keys you have revoked (only their hash, name and last four characters) until your account is deleted; the log of changes to keys and webhooks, 365 days; delivery records of webhook notifications, 30 days after the last attempt; stored answers to requests with an Idempotency-Key, 24 hours.
- KULTCOINS and purchase records of a deleted account are kept in pseudonymised form, because we need them for accounting, at most six years after the deletion; the KULTCOINS history of an account that still exists is kept as long as the account. Payment records (amounts, dates and the link to your customer record at Stripe) are kept as long as tax and commercial law requires, in Spain generally six years: we delete them six years after the payment or, for a subscription, after it ends, and the messages Stripe sends our server about payments after at most 365 days; Stripe keeps its own records under its privacy policy. This also applies to purchases in the apps (section 13): store subscription records are kept like payment records, at most six years after the subscription ends; the technical messages about them from RevenueCat are deleted after at most 365 days.
- Deleted accounts: a short record that an account was deleted (a masked e-mail address, the plan and the amounts), as proof of the deletion and for accounting, as long as the payment records.
- Welcome bonus and one-time promo codes: so that they cannot be claimed again with the same sign-in after an account is deleted, we keep a one-way keyed hash of its sign-in identifiers (for example of the e-mail address) for 3 years after the deletion. The address itself cannot be read back from it.
- Password reset links: a link you ask for in the sign-in window (2.14) works once and for 1 hour; a link an administrator creates for you when you write to support@kultai.agency from the e-mail address of your account works for 24 hours. We store only a hash (SHA-256) of a link, never the link itself, and delete that record 7 days after it was created; the e-mail with the link is removed from our outbox as soon as it has been sent. Setting a new password through the link signs out all your sessions.
- Requests from the website: as long as we need them to answer you and for our business relationship, and at most 24 months after the request was last updated. We delete them earlier on request.
- Subscription cancellations (2.3): 3 years from when we received them, as proof of the cancellation, also after the account is deleted; a request from the cancellation form whose e-mail address matches no account is deleted after 1 year.
- Service e-mails (2.14): the record of each e-mail (recipient, subject, status, times) 3 years, as proof that we sent the confirmations the law requires; its content and attachments 30 days after it was sent or given up, after that only a checksum (SHA-256) of the content, which lets us show which text was sent.
- Withdrawals (2.14): 3 years from when we received them, as proof, also after the account is deleted; a declaration whose e-mail address matches no customer, 1 year; submissions caught by the spam field, 30 days.
- Referral programme: codes, invitations, rewards and notices as long as the account exists; ambassador applications and views-bonus claims are deleted with the account, and the reward records linked to payments are kept like purchase records. A partner code remembered for a Telegram user who has not signed up is deleted after 60 days.
- Fraud-prevention records (accounts signed in from the same browser, self-referral checks): up to 12 months. The list of browsers that have signed in to your account (hashes of the known-device cookie, at most 20) is kept as long as the account exists.
- Reports (2.13): 12 months after our decision. If you delete your account, the note you wrote is deleted.
- Request check (2.13): the text of a refused request is not stored; refusals are counted only in memory, for 24 hours.
- Push tokens (section 13): until you switch notifications off, sign out on that device or everywhere, change your password or delete your account; also after three failed deliveries.
- Apple refresh token (section 13): until your account is deleted; we then use it to end the connection with your Apple account.
- Telegram bot: a draft you do not use is deleted after 30 days and the buttons under messages after 14 days; both are deleted at once when you delete your account. Support chat: the link between a forwarded message and your chat, 60 days; your support contact entry (Telegram ID, name, username, language) is deleted after 12 months without a message from you, except that we keep the ID of a person we have blocked from the support chat while the block lasts.
- Support assistant (2.8): conversations of signed-in accounts 30 days after the last message, then they are deleted (questions asked without signing in are not stored); support requests made through it, 12 months after they are closed; counts of topics, of questions without an answer and of ratings that are not linked to a person, without a time limit.
- Technical logs, including our web server’s access logs, are deleted after 30 days. API keys never appear in them in readable form.
- Cookies and browser storage: section 6.
- Backups are rotated and overwritten within about 90 days. Deleted data can therefore remain in a backup for up to about 90 days before it is gone for good.
8 Your rights
Under the GDPR you have the right to:
- access your data (Art. 15);
- have it corrected (Art. 16);
- have it erased (Art. 17);
- restrict its processing (Art. 18);
- data portability (Art. 20);
- object to processing (Art. 21), in particular processing based on legitimate interests;
- withdraw your consent at any time with effect for the future (Art. 7(3)); for cookies, through “Cookie settings” (section 6), and for push notifications, in your profile.
We make no decisions based solely on automated processing that have legal effects for you or affect you similarly significantly (Art. 22). The automatic check of requests (2.13) only refuses a single request; any restriction of your account is decided by a person. Personal offers (2.15) follow fixed rules and only add a lower price you may use; they are no such decision, and you can object to them.
To use these rights, write to us at support@kultai.agency. You can also delete your studio account yourself: Profile → Delete account. We reply within one month, as the GDPR requires, and may ask you to confirm that the account is yours. If your data reached us through a customer of our API, see section 14.
You also have the right to lodge a complaint with a supervisory authority (Art. 77). In Spain this is the Agencia Española de Protección de Datos (AEPD), www.aepd.es; you can also contact the authority in the country where you live.
9 AI content and other people’s data
- Your prompts, reference files and results are processed by the AI providers listed in section 4 in order to produce the result. Photos, videos and voice recordings you upload, including faces and voices, are therefore sent to kie.ai or Higgsfield in the USA and possibly to the developer of the model (sections 4 and 5). kie.ai’s terms allow it to use uploaded content to operate and improve its services, and Higgsfield’s API terms allow it to use content to train its models unless the customer opts out.
- No biometric identification. We do not use photos, videos or voice recordings to identify or verify anyone: we create no face or voice templates, compare no faces or voices and train no identity models of people. Talking videos, voiceovers, characters made from a photo and edits that keep a person’s look only generate new content from the files you choose. We therefore do not process biometric data for the purpose of uniquely identifying a person (Art. 9 GDPR). Photos and recordings can still reveal sensitive information, such as health or ethnic origin; we use them only to carry out the generation you ask for. Using the studio for facial recognition or biometric identification is not allowed (Terms of Use, section 16).
- Other people’s photos and voices. Before you upload a photo, video or voice recording of another person, you must have that person’s consent, both to the upload and to the use you intend for the result, and you must be able to show it. Do not upload such material without consent. The studio asks you to confirm this before you upload files, before you upload a face or a voice for a talking video and before you make a character or a Marketing image from a photo of a person, in the last two cases also that the person is 18 or older (Terms of Use, sections 6 and 16).
- Labelling. If you publish a realistic result that could be taken for real, in particular one that shows a real person or imitates a real voice, you must disclose that it was generated or manipulated by AI wherever the law requires it, in particular under Article 50 of the EU AI Act (Terms of Use, section 7).
- You are responsible for holding the rights to the files you upload.
- If someone has uploaded your photo or voice without your consent or created content that shows you, you can report it to us (Terms of Use, section 17) and also exercise your rights under section 8.
10 Children
KULT GENERATE is only for adults (18 and over). If you are under 18, please do not create an account. If you believe that someone under 18 has created an account, please contact us.
11 Security
Our server is in the EU, connections use HTTPS, passwords are stored only as salted hashes and API keys only as hashes, rate limits and abuse checks protect accounts, and databases and media are backed up every night; the copies kept outside our server are stored in Google Drive (section 4), and the secrets that sign webhook notifications are left out of them. Notifications to API webhooks are signed, and push messages to browsers are encrypted.
Your uploads and results are stored at long random addresses that are not listed anywhere and not indexed by search engines, so that the AI providers can fetch them. Anyone who has such an address can open the file, so share links to your files with care.
No system is completely secure, but we take care to protect your data.
12 Changes to this policy
We update this policy when the service changes, for example when new functions or new payment or sign-in methods are added, and publish the new version here with a new date. If a change concerns cookies that need your consent, the cookie banner asks you again.
Current version: 29 October 2026.
13 The apps, push notifications and in-app purchases
This policy also covers the KULT apps for iPhone and Android and the website installed on a home screen (a web app). The app shows the same studio; it adds the following processing:
- In-app purchases: purchases in the app are sold and processed by Apple or Google as the seller of record; we do not receive your card data. We use RevenueCat (RevenueCat, Inc., Florida, USA) as our processor to confirm purchases: it receives the store receipt or purchase token, the product, the price and a random app user id we assign to your account (not your name or e-mail). The purchase component in the app (RevenueCat’s SDK) connects to RevenueCat directly and also sends it the device type, the operating system and when you last used the app; like any connection, this also gives RevenueCat your device’s IP address. RevenueCat stores the data in data centres in the USA. Transfer: Standard Contractual Clauses under RevenueCat’s data processing agreement (section 5). When you delete your account, we delete your customer record at RevenueCat; we keep the records of your purchases as described in section 7.
- Push notifications (only if you switch them on in your profile and allow them on your device or in your browser; legal basis: your consent): we store the technical address of your device or browser (a push token or subscription) with its language and the time it was registered, and for a browser also the browser type. Through Apple Push Notification service, Google Firebase Cloud Messaging or the push service of your browser we send a short message that a generation is ready or did not work (for example “Ready: video” and the model’s name) with a link to it. If a storage period for results is set (section 7), we also tell you this way before your results are deleted. The message never contains your prompt, and messages to browsers are encrypted so that the push service cannot read them. You can switch notifications off at any time in your profile or in your device settings; how long we keep the token: section 7.
- Sign in with Apple and Google in the app work like on the website (section 2); with Sign in with Apple you can hide your e-mail address (Apple’s relay address). We keep Apple’s refresh token only to revoke it when you delete your account.
- Camera and photo library: the app accesses them only when you choose to attach or save a file; we receive only the files you upload.
- Sending your content to the AI providers: before the first generation, upload or chat message in the app, the app explains that your prompts and uploads are sent to the AI providers of section 4 and asks you to agree, as the app stores require. We store the time you agreed. The legal basis of the sending itself is our contract with you (section 3): without it we cannot create what you ask for. If you do not agree, the app sends nothing and cannot generate.
- Stricter checks: in the app, the automatic check of requests (2.13) also refuses explicit sexual requests, and chat answers are checked in the same way.
- The app contains no advertising, tracking or analytics SDKs.
14 The API: data you send for your customers
This section is for customers who use KULT GENERATE through its API (Terms of Use, section 18) and for the people whose data they send us.
- Roles. For the data about the API customer’s own account we are the controller (2.12). When an API customer sends us personal data of its own customers or end users, such as prompts, files, addresses of files or metadata, the API customer is the controller, and we process these data on its behalf as its processor (Art. 28 GDPR), only to carry out its requests: we store them in its account, send them to the AI providers, return the results and send the notifications it asks for. We do not use them for purposes of our own, except to protect the service against abuse and to keep the accounting records of the requests.
- Where the data go. Requests through the API run only at kie.ai (USA) and, through it, at the developer of the chosen model (section 4); they never run at Higgsfield, and the API offers no chat. Our server is hosted by Hetzner in Germany, and its backups are stored in Google Drive (section 4). We have no data processing agreement and no Standard Contractual Clauses with kie.ai or with the developers of the models (section 5). Before the API customer sends us its end users’ personal data, it must check that it may do so under these conditions and inform its end users.
- Addresses the customer chooses. If the customer gives us the address of a file, our server downloads the file from there. If the customer sets a webhook, we send the status of each request, its metadata and time-limited download links to the address the customer has chosen. These addresses are chosen by the customer and are not our recipients.
- Deletion. The data stay in the customer’s account until the customer deletes them or the account (section 7); backups as in section 7.
- Data processing agreement. If you need a data processing agreement under Art. 28 GDPR, contact us at support@kultai.agency before you send such data. kie.ai and the developers of the models are not bound to us by such terms, and kie.ai may use content to operate and improve its services (section 4).
- End users’ rights. End users should send their requests to the API customer. If they write to us, we pass the request on to the customer, where we can identify it, and help it to answer.