How we collect data
We collect information from registration and request forms, browser and server requests, the features you use, and the sharing you initiate. When you share content, the people you select or people with access to a public link may be able to see it. We disclose data to public authorities to the extent required by an applicable legal obligation.
Registration details, date of birth, country or region, and session and security records
Set up and operate an account as necessary for the contract.
Chats, uploaded files, Canvas or app content, groups, and sharing choices
Provide the selected service where processing is necessary for the contract.
AI prompts and relevant conversation context or a file, image, or audio input when the feature uses it
Carry out the AI action requested.
In-app mail and live voice information, including communications where the feature handles them
Deliver communications or run a session.
Forms responses and form versions, account name and email for forms requiring sign-in, and a form-specific identifier when repeat submissions are limited
Deliver responses to the form owner and apply the submission limit. The owner of a form that does not require sign-in receives the information entered in the response. We do not attach the account name or email to new responses to those forms.
Map queries, directions, and locations you choose
Find nearby places, resolve addresses, or calculate a route. The map provider can receive the location data needed for the request. People who receive a shared route link can view the locations it contains.
When you ask Map Assistant for an AI reply, we send your question, recent conversation and any available route, selected place and location information to the configured AI provider. If you use voice commands, your browser may send audio to a speech recognition service. You can also type your question.
Support reports, content notices, decisions, and appeals
Handle requests and protect the service, comply with legal duties, or establish and defend rights. The basis depends on the request and its purpose. If you report AI content as a guest, we link your description and any excerpt you choose to your guest record so you can check the outcome in the same session.
If you report a member’s message in a private Community conversation or AI Group Chat, the authorised review team receives your explanation, the message identifier and up to the first 1,000 characters of the selected message. The report excludes conversation history and attachments.
Cookies, browser storage, IP addresses, and request logs
Maintain sessions and preferences, operate and secure the service. See Cookies and Browser Storage.
External services
Depending on the feature, we may use the configured Amazon Bedrock or Evren AI service, Vercel AI Gateway for images, TomTom for maps, LiveKit where enabled for Community voice rooms, and the configured mail provider for transactional emails. If you ask Çevik Mail to review a thread with AI, the action shows you the recipient first. It can send the subject and excerpts from the last eight messages and, depending on the action, the sender’s domain, your question, or recent assistant exchanges.
In live conversations with ÇevikAI, your browser’s speech recognition service may process microphone audio. Depending on your browser and how the feature runs, ÇevikAI sends speech audio to the configured AI provider for transcription. We process your speech transcript and relevant context to answer. While camera or screen sharing is on, a frame may be sent with your question.
The current web infrastructure, application database, file storage, Çevik Mail and self-hosted LiveKit service run on our Oracle Cloud server in Frankfurt, Germany. Bedrock AI requests are sent to Northern Virginia in the United States. The model’s default data retention setting is active; we do not promise that requests and responses are never retained. Transfer safeguards and retention conditions for these overseas recipients require assessment for the services used.
Weather matches city and district names, search suggestions and country selections on the ÇevikAI server using GeoNames data. To retrieve forecasts, the ÇevikAI server sends coordinates truncated to at most four decimal places to MET Norway. MET Norway states that its API access logs are stored in Oslo, Norway. If the optional BigDataCloud server service is configured, it also receives coordinates to identify a place name. When that service is unavailable, the nearest settlement name is looked up in the GeoNames data on our server and displayed as an approximate location label. If no suitable settlement is found, coordinates are used as the label. You can search by city without granting location permission.
When you choose to load an external video on a Site Builder publication, your browser connects to the provider named on screen. The provider may receive connection data such as your IP address and the site’s origin. Other external images and media added by publishers may also cause your browser to request resources from their servers.
When you choose to open a map in the Maps app or a small search preview, your browser connects to TomTom. That connection can include your IP address and technical requests for the displayed map area. You can search in Maps without sharing your device location.
The general Web, Shopping and News search tabs use Fireball, Fynd, Wikipedia and Wikidata through ÇevikAI’s self-hosted SearXNG service. Queries, language and safe-search preferences may be sent to these engines. News search selects general web results using a news-oriented query and page filter; check dates and freshness at the source. Video search retrieves links through Fireball Videos, Pixabay Videos, SepiaSearch and Wikimedia Commons Videos. Loading video thumbnails connects your browser to the relevant media server; opening a link connects it to the publisher. These services may receive your IP address and connection information.
Data use in AI features powered by Evren
Evren states that its LLM service processes prompts, responses and usage records on its infrastructure in Türkiye and does not transfer them abroad.
At this stage, we cannot rule out the use of content sent to Evren for model training or development. We have not yet verified that this use is disabled for ÇevikAI’s account, how long content is retained, or which copies deletion covers. We will update this notice when it is clear how the data-use permission and right-to-be-forgotten mechanisms in the provider’s terms apply to our account.
Do not submit third-party information you are not authorised to share or confidential documents to these features. You can also ask us to review data sent to Evren through the request channels on the Data Rights page.
Account closure and deletion
When account closure starts, your active sessions end. You can restore your account within 30 days. Permanent deletion starts after that period if you do not restore the account. If ownership of a shared group, community, or list prevents closure, transfer ownership or contact us through Data Rights. Statutory erasure requests follow the conditions and time limits of the applicable law.
Search, news, maps, finance and stock media
General search queries, language and safe-search settings are sent through ÇevikAI’s self-hosted SearXNG service to Fireball, Fynd, Wikipedia and Wikidata. The engines used depend on the category and engine availability. These services may receive the query and our server’s connection information. We do not attach your account identifier or email address to the search request. Search history and saved bookmarks are processed separately in your ÇevikAI account; incognito search follows the existing history preference.
News comes from TRT Haber and HaberSitesi.org in Turkish, and the European Commission and News Central Site in English. The separate News app retrieves category feeds on our server; the search News tab applies a news filter to general web search. We do not send your account identifier or chat text to these publishers. For HaberSitesi.org RSS, we preserve the headline, short first paragraph and source link; images from this source are labelled illustrative. When an AI news summary is requested, selected headlines, short text and source information may be sent to the configured AI provider. Opening a source link or loading an external image connects your browser to the relevant publisher or media server. English news cards include the publisher name, source link and text licence.
In Maps, place/address queries, route points and device coordinates you choose to share are sent from our server to TomTom. Map tiles and traffic layers connect your browser to TomTom, which may receive your IP address and the displayed map area. The shared two-minute server cache for place search, reverse geocoding and routes has been removed. Favorites and map records you save remain separate account data.
The Market Tracker requests reference quotes from Frankfurter v2 and Gold API on the server using currency or metal codes. When you choose to load a TradingView chart, your browser requests the selected symbol and sends connection and device information to TradingView. Opening an external BIST or ETF link connects to the named provider. Market favourites and comparisons sync asset identifiers and display metadata in your ÇevikAI account.
Stock-media search uses the Openverse and Wikimedia Commons catalogues. The video, presentation and design editors retain their CC0/PDM filter. Loading a preview or asset may connect your browser to the record’s media server. A catalogue license label does not grant separate permission for personality, trademark or other third-party rights.
For image search, queries and page information are sent from our server to Openverse, Wikimedia Commons, the Art Institute of Chicago and Iconify. Lucide, Material Icons, Devicons and selfh.st icons are searched through SearXNG. Publisher artwork, creator and licence metadata are processed for eligibility checks. Some known Turkish visual concepts may be translated locally for catalogue search without an AI call. Loading previews or icons may connect your browser to the record’s media server, jsDelivr, Google Fonts/Gstatic or Iconify. These services may receive your IP address and connection information. Your account identifier and email address are not attached to catalogue queries.
Audius music catalogue and audio playback
Audius is the active provider for Music. Your search text and requested track identifiers are sent from our server to the Audius public API. We do not attach your account password, email address or private conversations to these music requests. Loading audio or artwork connects your device to content nodes in the Audius network; those nodes may receive your IP address and connection details and may be located abroad.
We do not offer audio downloads or offline archiving. Audius catalogue information is read for the respective request or session; persistent library and last-track records retain a track identifier representing your selection. Playlist names, likes, listening activity and sharing choices are your account data. Existing YouTube records are preserved; while that provider is disabled they do not play or trigger an automatic fallback to YouTube.
The Audius privacy policy is at https://audius.co/privacy, its API terms at https://audius.co/documents/ApiTerms.pdf and the Open Music License at https://openaudiofoundation.org/open-music-license.pdf. Artist access restrictions and the conditions applying to each track remain in effect. Catalogue availability or a verified-account badge alone does not establish ownership of every upload.
News images and source information
Images in TRT Haber RSS and HaberSitesi.org feeds may load in your browser from the publisher’s image addresses. HaberSitesi.org images are labelled illustrative. For European Commission and News Central Site feeds, we display text and source links; the right to display photographs is verified separately. These image servers may receive your IP address and technical connection information. Opening a story or image-source link takes you to that publisher. Image requests use a no-referrer policy to avoid sending the page address.
Source names, image credits and supplied licence or terms links are retained on news cards. A story’s text licence may not cover every photograph. Permission to reuse TRT images is a separate matter; RSS availability is not presented as permission for every use or as a guarantee that an image is copyright-free. Third-party image rights on News Central Site and other sources are assessed separately.
Music lyrics and search matching
When you open the lyrics section, the song title, artist when known, and recording duration may be sent through our server to LRCLIB (lrclib.net) to match lyrics. Your account identity, email address and conversation content are not sent for this request. We do not have a verified specific agreement covering the service’s retention of its connection and request records.
Artist information may come from provider fields or an explicit artist–song credit in an upload title; a title credit is not proof of verified identity or copyright ownership. Tracks without artist metadata remain in the catalogue. Automatic recommendations use the artist, genre and language context of recent listening.