Hominin

Privacy

Effective 2 October 2026 · Updated 10 October 2026: first-party analytics — while Search data collection is on, searches now keep your IP address and a visitor ID stored in your browser, and repeat visits are linked, to measure use and whether visitors return (first-party only, never sold; Private mode, and not accepting the banner where it is opt-in, keep nothing). Earlier, 8 October 2026: the Stack Exchange API, the crawl lists, server locations, the traffic-summary country, address-bar search counting, and privacy choices by country

Who we are

Hominin is a search and answer engine at hominin.com, run by Hominin LLC, 8 The Green, Suite B, Dover, Delaware 19901. For anything on this page, including a request to see or delete your data, write to hello@hominin.com. The Terms cover how you may use Hominin; this page covers your data.

Your choices

In Settings, under Privacy; Private mode is there too. Your choices are kept in your browser, and with your account if you are signed in, and our server applies them to every request, whatever page it comes from. Where you have not made a choice, the default depends on where you are. We work that out on our own server from your IP address, using that request only; we do not keep the result or the address. In the United States the defaults below marked “On (US)” apply. Everywhere else, or if we cannot place you, those switches are off until you turn them on.

On your first visit we show a notice saying which of these apply to you, with buttons to turn them off or use Private mode. Outside the United States, before collection or training starts we show what it means and ask you to turn it on; there we never take silence, or agreeing to the Terms, as that choice. Until this change on 2 October 2026 every switch started off for everyone; a switch you had set then keeps your setting. Turning a switch off stops future collection. It does not delete what was already kept: for that, see Your rights. Settings, Privacy, also has Clear history on this device, which removes recent searches, memories, interests and pages marked read, and leaves the pages and places you saved.

What we collect, and why

Running the service (always)

Our web server records each request: your IP address with its last part set to zero (so it points at a network, not a household), the time, the page requested without any search terms (for a search typed into your browser's address bar it notes only that the home page was opened with a search, never the words), the page you came from (again without search terms), and your browser's user-agent string. It is deleted after 14 days. We use it to keep the service running, stop abuse, and see how much the site is used, including which of our ad links (addresses starting hominin.com/go/) brought visits. Our administrators can see a summary of traffic by network from this log, with the country each network is in, which we look up from the shortened address in a table on our own server (from DB-IP.com, CC BY 4.0); nothing is sent anywhere to do it, and the country is not stored. Because the IP address is shortened, a network is usually an organisation, a mobile carrier or a neighbourhood rather than one person.

For each search we also record how long it took, how many results it found and what kind of search it was, without its words unless search data collection is on for you. These records are kept for 30 days. Our application log holds timings and errors; where it refers to a search it uses a short code that cannot be turned back into the words.

Until 2 October 2026 the application log also held the first 40 characters of most searches, including private ones. That has stopped. Entries written before then are still in that log until we clear it.

What you search for

To answer a search we look it up in our own index and, for some kinds of question, ask outside services (listed below). Answers are cached on our server for 45 minutes, in memory, except in Private mode. With Search data collection on, we keep, on our own servers and never sold:

• a count of how often each search was made, for 90 days after it was last made;
• searches our index answered badly, for 30 days, so we know what to crawl next;
• the words of each search beside its speed and result counts, your IP address, and a visitor ID we store in your browser, for 30 days;
• which results you open, how long the page was in view, how many searches you made, and whether you returned — linked by that visitor ID, so we can tell a returning visitor from a new one and measure whether people come back.

The visitor ID (hominin_vid) is a random value kept in your browser; it holds no name, email or account, is never shared with anyone, and is removed by “Clear history on this device”. In opt-in regions, rejecting the cookie banner, and Private mode anywhere, mean no visitor ID is set and none of the above is kept.

Whatever your settings, and in Private mode too, we count each day how many searches were made in Private mode, with collection or training on or off, from a region where it starts on or off, from a child's account, and with Global Privacy Control sent, in total and for each country. Each is one number per day. The country is looked up from your IP address on our own server at the moment of the search, and the address is not kept. The count has no search words, account, address or time of day in it, so it cannot be linked to you or to a search. It tells us whether these settings work as this page says.

If someone types personal details into a search, those words are personal data even without an IP address or account, and we treat them as such.

Improving Mira (on in the US until you turn it off; elsewhere only if you allow it)

With Use my content to improve Mira on, and never in Private mode, we keep for up to 90 days: the AI answers written for your searches with what they were written from (the sources, and, when your location was used, the nearby places or weather they drew on); your questions to Mira and her answers; questions you ask about a page in the reader, with that page's text; and answers you rate up. These copies carry no IP address or account. We use them to train and evaluate future versions of Mira on our own computers. A model that has already been trained on an example cannot unlearn it, so turning the switch off or asking us to delete stops future use but does not change models already trained. Copies made from this data for training are kept on our own computer; we are setting a time limit for them and will state it here.

Separately, short daily briefings written from public news and shared playlist topics are kept the same way; they contain nothing you typed.

Talking with Mira

• Mira is an AI, not a person, and can be wrong. Before your first chat or call we say so.
• To answer you, your browser sends your question and the last few turns of the conversation; the server uses them for that answer. They are kept only as described under Improving Mira, while that is on.
• Your chat history. If you are signed in and not in Private mode, each question, Mira's answer and the sources she cited are saved with your account, encrypted, so you can reopen the chat. Nothing else is saved with it. What is not encrypted is which account a chat belongs to and when it was written. The encryption key is kept on our server apart from the chats, and a copy is inside our encrypted configuration backup. You can delete one chat, or everything from the last hour, day or all time; pause saving; and have chats deleted after 3, 18 (the default) or 36 months, or not at all. Chat history is not used for training. Signed out, a chat is kept only in your browser tab.
• On a call, your voice is turned into text on our own computers and the recording is discarded, unless you tick "Also keep recordings of my voice" (it starts unticked). Then each recording is kept with its text for 30 days to help Mira understand speech, and deleted. A voice can identify the person speaking.
• Memories are kept in your browser only, and only while they are on; see Your choices.

Ratings and feedback

A thumb on an answer stores the rating, the question, the start of the answer, its kind, the page it quoted, the reason if you give one, and the date, for a year; not the time, your IP address or anything that says who you are. It goes into the training copy only if it rates up an answer our own model wrote and training is on for you. Feedback and reports about a page keep the day, the reason, your note, your search and what you reported, for a year. To stop one sender making hundreds, a scrambled form of your IP address is kept in the server's memory for up to a day; it is never written to disk.

Accounts

Accounts are optional and use passkeys: no password, no email address, no age. We store the handle you choose, a display name and avatar colour if you set them, the public half of each passkey, hashed one-time recovery codes, a signed-in session in a cookie named hominin_session (sent only to hominin.com, unreadable by page scripts, expiring after 30 days), and your privacy choices. Deleting your account (on My profile) deletes all of that, your saved items and collections, your Mira chat history, the preferred sources and notes you made, and your business claims with anything not yet approved, at once.

What you save

When you are signed in, what you choose to save (pages, pictures, videos, places and searches, with any notes you add and the collections you sort them into) is kept with your account so you can see it on any device you sign in on. Signed out, saving stays in your browser and nothing is sent to us. Signing in does not upload it: Saved & Collections offers an Import button, and only that sends it.

On our server it is encrypted (AES-256-GCM) with a key kept apart from the databases. Our database backups do not contain the key; it is backed up only inside a separately encrypted archive whose decryption key is not on any of our servers. Each account's items are sealed with a key derived for that account. The address, title, description, place details, the saved search, your notes and collection names are all encrypted. Left readable are only the account number, the kind of item and the dates. This is encryption on our side, not end-to-end: our server decrypts it to show it to you, and someone who controlled the server could read it. It protects the copies in our backups and a copied database file.

It is kept until you remove it or delete your account, and it is in Download my data. Copies in our backups are gone within 7 nights of being deleted. Saving is not consent to anything else: what you save does not change your results, is not used for personalization, and is not used to train Mira. Your saved items and collections are private; nothing is shared or made public, and there is no way to do so yet.

Stored in your browser

When you are signed out, saved pages, places and searches, saved sessions, settings and your privacy choices are kept in your browser's storage, and so, outside Private mode, are recent searches, pages you read and a short answer cache. They stay on your device. Interests, your saved place and memories are sent with searches only as described under Your choices.

Location

• Your IP address. When a search is about what is nearby, or about the sun where you are, and your browser has not given us a position, we look the address up in a city table on our own server (IP to City Lite from DB-IP.com, CC BY 4.0) and answer for that city. The address is used for that search and dropped: not stored, logged or sent to anyone.
• Your browser. If you allow precise location, your browser gives us a position, used for nearby results, weather and maps, and not stored.
• A place you keep in this browser (Change, under local results) is sent with a search that has no position, if Personalize results is on and you are not in Private mode, and is not stored by us.

Permission to use your location for a search is not permission to keep it: location goes into a training example only as the text of an answer, and only while training is on.

Reviews and corrections

Under a place you can write a review or report a problem. We keep the place, what you wrote and the date, and nothing that says who you are; you do not need an account. A review is public. Reviews with links, emails or phone numbers are refused, some are held for us to read first, and one reported by three readers is hidden until we look. Because a review carries nothing that says who wrote it, write to us with the place and the text to have yours removed.

Searching with a picture or a document

It is sent to our server, read in memory, and discarded: not stored, logged, used for training or sent anywhere else.

Businesses, sponsored results and the API

If you claim your business's website we keep the site, which account claimed it and how it was verified (our server visits your home page and one file, and reads your DNS records). Logos and photos you upload are stored without location and camera details, reviewed by a person, and shown publicly once approved; you can remove them or your claim at any time. Sponsored results are chosen from the words of the search alone; we count how often each is shown and clicked, and nothing about who. So that you do not see the same one over and over, your browser keeps on your device a list of the sponsored results it showed you in the last day and skips one already shown three times. That list never leaves your device, and is kept only while Search data collection is on. API keys are counted against a daily limit.

Notifications

If you turn on notifications we store the push address your browser gives us, which identifies your browser to its push service. Turning notifications off in your browser stops them; write to us to have the address removed.

Who else sees data

Services we ask on your behalf

For some searches our server sends the words of the search, never your IP address or an identifier of yours, to outside services, whether or not you are in Private mode: Wikipedia, Wikidata and Wiktionary; the Internet Archive; Open Library; MusicBrainz and Apple's iTunes search; YouTube (the YouTube Data API, for video and live-stream searches); eBay; OMDb; TheSportsDB; MET Norway for weather; OpenStreetMap's Overpass and OSRM for places and routes; IsThereAnyDeal for game prices; PoetryDB; Hacker News (through Algolia), Lemmy (lemmy.world) and Mastodon (mastodon.social) for discussions under a search that asks what people think, never for a private search; Twitch, for live streams; IGDB and speedrun.com, sent only a game's name; Steam's store, sent only a game's Steam number; a domain's registry over RDAP, sent only the domain name; MyMemory (Translated srl, Italy) for translations, sent the text to translate; and the Stack Exchange API, for programming questions, to find matching Stack Overflow questions to link to, never for a private search. We do not send searches to other web search engines.

Links to Reddit, Stack Overflow, carriers' tracking pages and other sites are ordinary links: nothing goes to them unless you follow one. Site icons beside results are fetched by our own server from each site and served from ours.

Public lists and indexes we download

To decide what to crawl, our server downloads public lists and looks sites up in public indexes: the Tranco ranking of popular websites, and Common Crawl's index of the pages it has found (asked by a site's name only). Some of the sites looked up are ones that searches came back thin for, but nothing about you, and not the searches themselves, is sent.

Game server lists

To show game servers in search results, our servers download the public server lists that FiveM and RedM (Cfx.re), open.mp, Multi Theft Auto, Vintage Story (Anego Studios), BeamMP (BeamNG.drive), TruckersMP, Northstar (Titanfall 2) and Factorio (Wube Software) publish, about every half hour; Factorio's is requested with our own factorio.com account. For QuakeWorld and Zandronum (Doom), our servers ask the games' master servers for their lists and then ask each listed game server for its status, as the games' own server browsers do; those replies include the names of the players on each server, which we count and don't keep. These requests are made by our servers, not your browser, and contain nothing about you or your searches. We keep each server's name, address, player count, tags and settings, and never the names of the players on it.

YouTube

Hominin uses YouTube API Services. By using the YouTube videos and players on Hominin you agree to be bound by the YouTube Terms of Service; Google's handling of information is described in the Google Privacy Policy. We don't use Google sign-in or access any YouTube account. Video results for a search (titles, channels, links, thumbnail addresses) are cached on our server for up to 7 days, not linked to you. Players load in privacy-enhanced mode (youtube-nocookie.com); YouTube sees your IP address and may store data on your device once you play a video.

Services your browser contacts directly

Pictures, thumbnails, pronunciations and sounds load from the sites that host them, most often Wikimedia Commons; game art from Steam and IGDB; map tiles from OpenFreeMap; video players from YouTube and Twitch when you open them; and, if you share your location, OpenStreetMap's Nominatim and OpenTopoData receive your coordinates. Each sees your IP address as any website would.

Where it is processed

Hominin runs on servers in Finland (search) and Germany (the website's front end and storage). AI answers, Mira's voice and speech-to-text run on computers we operate in the United States, reached over encrypted connections; the question, its sources and any context sent with it go there to be answered. Every night our databases are copied to a storage server in Germany (the last 3 nights kept), the largest directly from the search server in Finland, and from there to a computer we operate in the United States (the last 7 kept). Our configuration and keys are copied encrypted. The database copies on the storage server are not encrypted at rest. Data deleted from our live systems remains in these copies until they age out, at most 7 days.

How long we keep it

Your rights

• Search without an account, use Private mode, and turn collection, training, memories and personalization on or off at any time.
• Clear what is stored in your browser from Settings, Privacy, or by clearing site data for hominin.com.
• Download a copy of everything kept with your account (Download my data, on My profile), and delete your account, saved items, collections, chats, preferred sources and notes yourself.
• Ask us at hello@hominin.com for a copy of the data we hold about you, to correct it, or to delete it. We reply within 30 days. Training examples and ratings carry no identifier, so we find those only from their words; searches, while collection is on, carry your IP address and a visitor ID, so give us the exact text and roughly when (and your IP or visitor ID if you have it) and we will delete matching entries.

If you are in the European Union, the EEA or the United Kingdom: search data collection, training and memories are off until you turn them on, and we rely on your consent for them (including the IP address and visitor ID kept with your searches), for precise location and for notifications, which you can withdraw in Settings at any time. We rely on our legitimate interest in running, securing and measuring the service for the access log and the operational records above. You may object, and complain to your data protection authority (in the UK, the Information Commissioner's Office). If you are in India, you can raise a grievance with us at the same address.

Children

Hominin is a general search engine for a general audience; it is not directed at children, and some of what it shows (games, drawing, explainers) may still interest them. We do not ask anyone's age, do not guess it from searches, and do not offer accounts or features for children under 13. Because we cannot tell a child from an adult when nobody is signed in, the defaults in Your choices apply to every visitor: in the United States the words of searches are kept as described above unless the switches are turned off or Private mode is on, and we recommend Private mode on a device a child uses. Elsewhere a visitor with the default settings leaves only the access log and our operational records. If we learn that an account belongs to a child under 13, its optional collection, training, memories, chat history, kept recordings and business features are switched off, and we delete it. A parent or guardian who believes their child has used Hominin can write to us and we will delete what we can find.

Security

Connections to hominin.com are encrypted. Access to our servers and administrative tools is restricted to named keys and addresses.

Changes

When we change what we collect, the choices you have, or how long we keep anything, we update this page and the date at the top, and show the first-visit notice again when what happens to your searches by default changes for you. A switch you have set keeps your setting through a change. On 2 October 2026 the defaults in the United States changed from off to on. On 10 October 2026 Hominin moved to first-party analytics: while Search data collection is on, searches now keep an IP address and a visitor ID and repeat visits are linked, to measure use and whether visitors return — first-party only, never sold, off in Private mode and, where collection is opt-in, until you accept the banner. A new kind of optional collection will not start without that notice.