Privacy Policy

Who we are

Rumore is operated by Inevitable AI ("Rumore", "we", "us"). Rumore is a reputation and customer-communication platform for owner-run local businesses: it gathers reviews, messages, listings and social activity into one place and helps owners respond, with AI assistance that the owner approves.

For anything in this policy, write to us at privacy@rumore.ai.

The short version

We collect only what we need to run the product.
We never sell personal data. We never use your business’s data, or your customers’ data, for advertising. We do use an advertising tracker on this website, for our own ads. In the EU and the rest of the EEA, the UK, Israel, Brazil, Korea, China, Saudi Arabia, Nigeria and Quebec it waits for your permission; everywhere else it runs until you turn it off, which is one tap from the banner or from Cookie settings in the footer.
Data from Google and Meta is used solely to show your business its own reviews, messages and comments, and to publish the replies you approve.
If you connect an AI assistant such as ChatGPT or Claude, it gets the data it asks for about the Location you chose, under its own provider's terms. More below.
Everything is stored encrypted in the European Union.
You can ask us to delete your data at any time.

Data we collect

Visitors and the waitlist

If you join the waitlist we store the email address you give us, and the business name and type if you add them. We use them to contact you about early access and nothing else.

We use three analytics tools to see how the site is used: Google Analytics, Microsoft Clarity and PostHog. All three set cookies, and we do not pretend otherwise; they are in the table below. PostHog also keeps the same anonymous reference in your browser’s storage, so it can tell that a visit today and one next week came from the same browser. We do not sell what any of them collect.

PostHog also records two things that are about our software rather than about you. The first is an error report when something breaks in your browser: what broke, the line of our code it broke on, and the page you were on, with any sign-in, reset or poster link stripped out of that address before it goes. The second is how quickly a page appears on your screen and how soon it answers your first tap. We use them to find faults and slow screens. Neither one carries anything you typed.

Once you are signed in, PostHog is also told which Rumore account and which business the activity belongs to, as the internal reference codes we use in our own database. That is so we can see how one business gets on rather than a heap of unattached sessions. It is never told your name, your email address, your business name, or anything at all about your Contacts. On our public pages, where nobody is signed in, none of this applies. If you allowed analytics before you signed up, the pages you looked at on this browser beforehand are joined to your account too, so we can see which of our pages led to it.

We also run one advertising tracker: the Meta pixel, for our ads on Facebook and Instagram. It is the reason we can tell which ad brought someone here rather than guessing, and it lets Meta show our ads to people who look like the owners already using Rumore. Where the law asks us to get your permission first, in the EU and the rest of the EEA, the UK, Israel, Brazil, Korea, China, Saudi Arabia, Nigeria and Quebec, it stays off until you accept. Europe is not one rule here: Switzerland, Turkey, Ukraine and the other non-EEA countries sit with the rest of the world below. Everywhere else, including the United States, it runs from your first page unless you turn it off. In the United States a banner says so on arrival, and pressing “Only essentials” turns it off. In the other countries where it runs by default there is no banner, and the way to turn it off is the Cookie settings link in the footer of every public page, which opens the same controls. There are also three screens that hold the banner back because it would cover the questions they ask (the quiz, and the two setup interviews); the same Cookie settings control is on those, in the bar at the top. The cookie settings let you keep analytics and refuse advertising specifically, and you can change your mind at any time. If your browser sends a Global Privacy Control signal, advertising is off from the start wherever you are, whatever the banner says and whatever you pressed before. The pixel sets the two cookies at the bottom of the table below, and we have turned off Meta’s automatic collection, so it sends only the events we name: a page view, which article, comparison, feature, industry or glossary page you read, running a scorecard, seeing a scorecard result, sending a message in the chat bubble, joining the waitlist, creating an account, and, once trials open, clicking through to start one. It runs on our public pages and on the signup form, and nowhere else: never on a signed-in screen, so nothing about the business you run inside Rumore reaches Meta, and never on a page whose address carries a sign-in or reset link. Joining the waitlist counts once per visitor however many times you ask.

None of those events carries the business name you searched for. Meta is told that a search happened, never what you typed, which is the same line we draw with our own analytics. The name used to sit in the page address while you read your scorecard, and Meta’s pixel reads the address of every page it runs on, so it travelled without our meaning it to. A scorecard now carries an opaque reference code for the listing instead, so the card still has a link you can share and the name stays here.

Three conversions are also reported to Meta from our own servers rather than from your browser: joining the waitlist, creating an account, and the trial actually starting, which is a different moment from the button click above. An ad blocker otherwise turns a real signup into a missing one, and we would be guessing again. The two that also happen in your browser, joining the waitlist and creating an account, carry the same reference number as the browser’s report when there is one, so Meta counts one signup rather than two. The trial starting happens on a page the pixel never runs on, so that one is sent only from us. All three reports carry your email address hashed rather than in the clear, so Meta can match it against an account it already has without us handing over the address itself. Hashing is not anonymising, and we will not pretend it is: anyone holding the same address can produce the same hash, which is exactly how the matching works, so treat it as your address in a different alphabet. Each report also carries the IP address and browser the request came from, which is what Meta uses to tell one person from another, the page you were on with any query string stripped off, and the pixel’s own cookies if you have them (the two named in the table above). Nothing else: never a Business’s own data, and never anything about its Contacts. All of it is sent only for a visitor whose advertising is on, whether they switched it on or their country leaves it on until they say otherwise. Refuse advertising and nothing goes, from the browser or from us.

Microsoft Clarity is session replay, which means it records how a page is used: where you click, how far you scroll, where you get stuck. We use it to find the parts of the product that confuse people. On every signed-in screen it is set to mask all text, so the recording shows the shape of the page and not the words on it, which keeps your Contacts’ names, phone numbers and messages out of it. That masking is enforced in our code, not in a setting someone could switch off by accident.

Which of these run is your choice, and you can change it whenever you like from Cookie settings at the bottom of this and every other public page, or from the account block once you are signed in. Where the law asks for your permission first, we set no analytics cookie and send no identifier until you give it: Clarity and PostHog do not load at all, and Google’s tag loads in a mode that stores nothing and cannot recognise you. If your browser sends a Global Privacy Control signal, advertising stays off without you having to do anything else, and it stays off even if you accepted cookies here before you turned that signal on, because the signal is the more recent word. It does not switch analytics off over the top of a choice you made yourself in the panel; that one is yours.

Cookies we set

Strictly necessary cookies are needed for the site to work or for something you asked for, such as signing in or a chat you opened, so they need no consent. Analytics and advertising cookies follow your cookie settings, as each row says. Under each description is the cookie’s kind, how long it lasts, and whose it is.

Cookie	What it does	Kind	How long	Whose
`rumore_session`	Keeps you signed in.	Strictly necessary	7 days	Ours
`rumore_device`	Remembers that this browser is one you signed up on, or confirmed with a code from your email, a passkey or a link we emailed you, so signing in here with your password does not ask for a code every time. It holds a random token, and we keep only a scrambled copy of it.	Strictly necessary	30 days	Ours
`rumore_login_challenge`	Ties the sign-in code we email you to the browser where you typed your password, so the code works nowhere else. Set only when we send you that code.	Strictly necessary	30 minutes	Ours
`rumore_webauthn`	Holds a passkey sign-in, or the setup of a new passkey, while it is in progress.	Strictly necessary	5 minutes	Ours
`rumore_oauth_staterumore_oauth_pkce`	Protect the trip to Google and back when you sign in with Google or connect a Google account, so nobody else can finish that trip in your name. Set only when you start one.	Strictly necessary	10 minutes	Ours
`rumore_google_hint`	If you sign in with Google and have no Rumore account yet, keeps that Google email address so that creating an account can offer the same Google account instead of asking you to pick again. Deleted as soon as anyone signs in or out on this browser.	Strictly necessary	30 minutes	Ours
`rumore_invite_pick`	Used when a business owner asks you to connect their Google listing for them: keeps the permission you gave Google tied to your browser while you choose the listing.	Strictly necessary	30 minutes	Ours
`rumore_calendly_staterumore_calendly_pkcerumore_square_state`	Protect the trip to Calendly or Square and back when you connect your booking calendar. Set only when you start that connection.	Strictly necessary	10 minutes	Ours
`rumore_oauth_interactionrumore_oauth_interaction.sigrumore_oauth_resumerumore_oauth_resume.sig`	Set only when you connect an AI assistant such as ChatGPT or Claude to Rumore. They keep that one connection request together while you sign in and choose a Location. Each one ending in .sig is a signature that proves the one before it was not tampered with.	Strictly necessary	30 minutes	Ours
`rumore_oauth_sessionrumore_oauth_session.sig`	Also only when you connect an AI assistant: remembers that you signed in for that connection, with its signature beside it.	Strictly necessary	1 hour	Ours
`NEXT_LOCALE`	Remembers the language you chose.	Strictly necessary	1 year	Ours
`rumore_consent`	Remembers your cookie choice, so we stop asking.	Strictly necessary	6 months	Ours
`rumore_consent_default`	Records your country’s starting position for analytics and for advertising, before you have chosen anything, so this site and our servers apply the same one. It holds no choice of yours and is replaced by your choice the moment you make one.	Strictly necessary	6 months	Ours
`rumore_wcs`	Keeps a chat you start with us in one thread, so a reply comes back to the right place. Set only if you open the chat window, never just for visiting.	Strictly necessary	30 days	Ours
`rumore_rec`	Remembers which business sent you here, if you arrived through another business owner’s personal link, so we can give you both the free month we promised. It names that business, never you, and it goes nowhere else. Set only if you follow one of those links, never just for visiting.	Strictly necessary	90 days	Ours
`rumore_src`	Remembers where you first arrived from, so we know which pages, answers and ads send people to us. If you arrived from an advert it also holds a reference to your individual click on it, which the ad network can match back to you. Never changed after that first visit. Set only while analytics is on. Where the law asks for your permission first, that means only after you give it.	Analytics	90 days	Ours
`rumore_signup_event`	After you sign up with Google, tells the page you land on to count one signup in our analytics, and that page deletes it straight away. It holds only the word “google”, nothing about you. Set only while analytics is on. Where the law asks for your permission first, that means only after you give it.	Analytics	10 minutes	Ours
`_ga_ga_<id>`	Google Analytics. Counts visits and pages. Set only while analytics is on. Where the law asks for your permission first, that means only after you give it.	Analytics	2 years	Google
`_clck_clsk`	Microsoft Clarity. Ties a session replay together. Set only while analytics is on. Where the law asks for your permission first, that means only after you give it.	Analytics	1 year, 1 day	Microsoft
`ph_<id>_posthog`	PostHog. Recognises your browser from one visit to the next, so a path through our pages to signing up reads as one journey. A copy is kept in your browser’s storage too. Set only while analytics is on. Where the law asks for your permission first, that means only after you give it.	Analytics	1 year	PostHog
`_fbp_fbc`	Meta pixel. Connects a signup back to the ad that led to it. Set only while advertising is on. Where the law asks for your permission first, that means only after you give it.	Advertising	90 days	Meta

We ask again every six months, and straight away if we change what any of this does.

The free reputation scorecard

The scorecard is open to anyone. There is no account to create and nothing to hand over: the whole result shows as soon as you search. That search sends the business name you type, and the approximate area you are searching from, to Google to find the listing, the same search the review link and QR code tool runs. To know whether the tool works and whether our ads reach the people we hoped to reach, we keep our own record of a few things when you use it: that a visit happened and on which page, whether the search produced a result, and the short visitor code described below.

These records use no cookies and do not need you to be signed in. To count people rather than clicks, we turn your IP address and browser into a short code using a secret only our server holds. Nobody who gets hold of our database can turn that code back into you, and it is regenerated every night, so it never grows into a history of your visits. We keep visit records for up to a year.

We never store your IP address or your browser in these records, only the code we make from them. Two other places do hold an IP address. Our servers keep a request log, which records the method, the path, the response and the headers of each request, and the headers carry the address it came from. We read it to run the service and find faults. The other is the short-lived counter that keeps the free tools from being abused, which is kept per IP address, holds a count and nothing else, and expires no more than a day after it starts. So there is nothing in these records to match a request from you against. That is why we cannot pull up your scorecard visit records even when you ask us to. Those records are deleted on a timer instead. Anything tied to a Rumore account is different, and we do delete that on request.

When a result has written reviews, we send the text and star rating of up to five of its public Google reviews, without reviewer names, and the business name to our AI provider to find what they have in common. We keep only the themes it returns, for a day.

Until September 2, 2026 the scorecard asked for an email address before it showed the full result. It no longer asks. If you gave it your address before then, the address sits on the same early-access list as the signup forms elsewhere on this site, together with the business name and type you searched for. It stays there until you ask us to delete it. There is no timer on this one. Write to privacy@rumore.ai and we will take it off. The address is not joined to the visit records above.

For those earlier results we also kept a copy of what you saw beside your address: the business name, the score, the star rating, the review count, where you ranked, the language you were reading in and your browser’s time zone. We delete that copy on a timer, 400 days after you opened it. We never sent the follow-up emails this page used to describe.

We are about to start keeping more, so here is what changes before it does. Soon these records will also carry the campaign label on the link you followed, where you arrived from as a category (an ad, a search, another site, or direct) rather than an address, whether the screen was phone-sized or desktop-sized, roughly how long the page stayed open, how many characters you typed, and whether you picked a suggestion or typed it out. We will take campaign labels only, never the per-click tracking codes ad platforms append, never the address of the page that sent you, and never your browser version. We will also keep the business name you search for, for 30 days, so we can find the searches that come back empty and fix them. That one will be stored on its own, dated only to the day, with nothing attached that could tie it back to you. When this starts, this paragraph moves to the present tense and the date at the top of the page changes with it.

The other free tools

The review link and QR code tool sends the business name you type, and the approximate area you are searching from, to Google to find the listing, and builds the link and the code in your browser. We keep nothing from it.

The review reply writer sends the review you paste, its star rating, the business name and the optional details you add to our AI provider, which writes the draft and does not keep the text or train on it. We do not store the review or the reply. To keep the free replies free for everyone, we count how many drafts come from the same IP address in a day. That count holds no text and expires a day after it starts. The same applies to the reply the scorecard writes under one of your reviews.

Business account holders

When you create a Rumore account we store your name, email address, password (hashed, never readable by us) and your business details: locations, opening hours, contact details and the profile information you enter. When you connect an external service (Google, Facebook, Instagram, WhatsApp, your booking or CRM tool), we store the access tokens that connection needs, encrypted at rest.

We also keep a record of activity in your account: sign-ins, the actions you take, and errors your browser reports back to us, which include the page you were on when something broke. We use them to support you and to fix faults. Sign-ins and the actions you take are kept for a year. Faults our own servers record are cleared within a week. The reports your browser sends have no automatic expiry yet, so they stay until you close your account, which clears all of it.

Your customers

To do its job Rumore processes data about your customers on your behalf: contact details you import or collect, appointment records from connected booking tools, the messages they exchange with your business, and the reviews they leave publicly. For this data your business is the controller and Rumore is the processor. We act on your instructions, and we support export and erasure for any contact record.

Google user data

This section covers everything Rumore receives from Google APIs. It applies when you sign in with Google or connect a Google Business Profile.

What we access

Sign-in (scopes openid, email): your Google account email address and account identifier, used only to create and sign you into your Rumore account.
Business Profile connection (scope https://www.googleapis.com/auth/business.manage): the list of business locations your Google account manages, each location's public details (name, address, phone, opening hours), the reviews on those locations (star rating, review text, the reviewer's public display name and photo, and timestamps), and the owner replies published on them. Google offers no narrower scope for reading and replying to reviews, which is why this one is requested.

How we use it

We show your reviews inside your Rumore dashboard, notify you when a new one arrives, and publish the owner replies you write or approve back to your Google Business Profile. Our AI assistant reads the text of a review to draft a suggested reply for you. A draft is only a draft: nothing is published until you approve it.

How we store it

Reviews and location details are stored in our database, hosted in the European Union. The OAuth tokens Google issues are encrypted at rest and are never written to logs.

How long we keep it

We keep the credentials for as long as the connection is active. Disconnect Google Business Profile and that connection's copy of the credentials is overwritten on the spot, while the reviews we already pulled stay as your own record. The access you granted us when you connected is held once for your whole Rumore account, and it outlasts the disconnect: that is what lets you reconnect in one tap instead of going through Google's consent screen again. Closing your Rumore account is what removes it, and there is no button for it before that. To cut our access at the source in the meantime, remove Rumore from your Google account settings, which is the one step we do not do for you today.

Who we share it with

No one, beyond the subprocessors that run the service itself: our EU hosting provider stores it, and a third-party AI provider processes review text transiently to draft a reply. The AI provider receives the text only to generate that draft, does not retain it once the draft is returned, and does not use it to train its models. We never sell Google user data, never use it for advertising, and never transfer it to data brokers or analytics products.

The one exception is an AI assistant the owner connects themselves, which can ask for the reviews of the Location they chose, along with any directory the owner connects it through. AI assistants you connect below explains what each receives and how to stop it.

Rumore's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Google user data is never used to develop, improve or train generalized AI or machine-learning models.

Meta data (Facebook and Instagram)

When you connect a Facebook Page or a linked Instagram professional account, Rumore stores the Page and Instagram account identifiers and names, the encrypted access tokens for the connection, the messages your customers send your business through Messenger and Instagram, and the comments on the posts you manage through Rumore. For a comment we store the commenter's public name, their account identifier on the platform, and a link to their public profile picture (the link, not a copy of the photo). Those three are what let the thread show who said what, and often we hold none of them. Facebook withholds a commenter's identity until Meta approves us for the permission to read comments, and it has not. So no comment reaches us with a picture today, and many arrive with no name either. Instagram never sends a picture at all. We use this data to show those conversations and comments in your inbox, to let you reply, and to let our AI draft replies for your approval. We do not read anyone's personal timeline, friends or private profile data.

Meta data follows the same rules as everything else here: encrypted EU storage, no selling, no advertising use.

Disconnect a Page or an Instagram account and two things happen immediately, not on a timer. The Social connection's copy of the access token is overwritten, and the commenter details on every comment we hold from that account are erased: the name, the account identifier and the profile picture link all go. That erasure is permanent, and reconnecting later does not bring them back. Direct messages run on separate connections that stay live until you remove them under Connections too.

What stays is your own record of the exchange: the comment text, your reply, and the Messenger and Instagram conversations. We keep those for as long as your account is open. Deleting your account removes all of it, and deleting your data below sets out the few things that do not follow that path.

If you are a Facebook or Instagram user and want data about you removed from Rumore, see our data deletion page. Removing the Rumore app from your Facebook settings sends us a deletion request too, but that request identifies you by an account id we cannot match against anything we store, so it does not on its own tell us which records are yours. Email us as well and we will complete it within 30 days.

How AI is involved

Rumore uses AI to draft review replies, message responses and social posts from the content your business already handles (a review's text, a customer's message, the photo on a social post, your business profile). Drafts are suggestions for the owner, and the owner decides what gets published. The AI providers we use process this content only to produce the draft. They do not keep it and do not train models on it. Our AI Policy explains this in full.

What we keep of an AI draft

When the AI drafts a reply to a customer message or to a social comment, we keep that draft from the moment it is written, whether or not you send anything, and we keep your final version next to it when you do send. That is so future drafts can sound more like you. We do not keep a separate AI draft of a review reply. We do keep the reply you send, and where a review site will not accept a reply through us we hold it so you can post it yourself.

All of it stays inside your own business and is never shared across businesses. Messages go when the conversation they belong to reaches your retention window, which is a year unless we have agreed otherwise with you. Social comments and reviews have no separate clock: they stay until you delete your account.

Where your data lives and how it is protected

All production data is hosted in the European Union. Data is encrypted in transit and at rest, OAuth tokens and other secrets carry an extra layer of application-level encryption, and access inside Rumore is scoped so that each business can only ever see its own data.

Sharing and subprocessors

We share personal data only with the vendors that run the service: EU-region hosting and database infrastructure, an email delivery provider for the messages you ask us to send, our payment provider for billing, and a third-party AI provider for drafting. Each one is bound by a data processing agreement. The AI provider processes the content only to produce a draft, does not retain it afterward, and does not use it to train its models. Nobody pays us for personal data, and none of these vendors may use it for their own purposes.

There is one exception to that last sentence, and it is the whole reason this paragraph exists rather than a blanket promise: Meta. While advertising is on for you, the Meta pixel and the two reports we send from our own servers hand Meta identifiers about your visit, and Meta may use them to build an audience of people who resemble the owners already using Rumore. Under California law that counts as sharing for cross-context behavioural advertising, whether or not money changes hands, so this page says it plainly and leaves the choice to you. Since September 20, 2026 it happens on our public pages unless you turn it off, which is the choice California law asks us to offer rather than a yes it asks us to wait for. One click from Cookie settings, or the banner where one is shown, ends it, and a browser sending Global Privacy Control never starts it.

The four measurement vendors above (Google Analytics, Microsoft Clarity, PostHog and Meta) are subprocessors too. Where the law asks for your permission first, none of them stores anything or identifies you until you give it, and everywhere else you can turn them off from the same place. Advertising follows that same rule now: off until you allow it where permission is required, on until you refuse it everywhere else. Microsoft Clarity is the one that reaches furthest inside the product, because session replay covers the signed-in screens and not only this site, so it is named in the Terms of Use along with the masking rule that applies to it. Meta is the one that reaches furthest outward, because it is an advertising network rather than a vendor working only for us, so it is named there as well. If we take on a new subprocessor that handles data from a Rumore account, we tell account holders before it starts and they can object.

AI assistants you connect

An owner can connect Rumore to an AI assistant of their own, such as ChatGPT or Claude. Nothing reaches an assistant unless the owner connects it: they sign in to Rumore from inside the assistant, or create a key in Rumore, and choose the one Location it may use. After that, when the assistant asks Rumore for something, we send it what it asked for about that Location. That can be reviews, conversations, listings and settings, and the details of any Contact a request touches, such as their name, phone number, email address or the messages they sent the business.

An assistant connected by signing in can also act for the owner, for example post a review reply, send a message or a Review Request, or publish a Post. It is told to describe each action to the owner and wait for their yes in the chat before it asks Rumore to do it, and Rumore still applies its own rules (it never messages a Contact who opted out, for example). Each action then shows in Rumore under the assistant's name. An assistant connected with a key can only read: it cannot send, post or change anything.

The company that runs the assistant (OpenAI for ChatGPT, Anthropic for Claude) is not one of our subprocessors and has no agreement with us about this data. It handles what it receives under its own terms and privacy policy, not ours, and it may process that data outside the European Union. What it already received stays with it, and we cannot delete it there.

An owner can also connect through a directory that relays the connection, such as Smithery. That service then sits between the assistant and Rumore. Everything the assistant asks for or sends about the Location passes through it, and it may keep the connection's access so it can reconnect. Like the company that runs the assistant, it is not one of our subprocessors, it handles that data under its own terms and privacy policy, and it may process it outside the European Union. What it already received stays with it, and we cannot delete it there. Connecting straight from the assistant keeps it out of the path.

To stop a connection, open Settings in Rumore and revoke the connection under Connected assistants, or revoke the key. The assistant, and any directory relaying it, loses access on its next request. You can also remove Rumore from inside the assistant, but revoking in Rumore is the one way that is sure to cut off a relay too.

Your rights

Under the GDPR (and similar laws elsewhere) you can ask for access to the personal data we hold about you, ask us to correct or delete it, ask for a portable copy, and object to or restrict certain processing. Email privacy@rumore.ai and we will respond within 30 days. If you are a customer of a business that uses Rumore, we may refer your request to that business, since it controls your data, and we will help it respond. You can also complain to your local data protection authority.

Deleting your data

Business owners can disconnect any integration from inside Rumore. That overwrites our copy of that connection's credentials on the spot, and for a Facebook Page or Instagram account it also erases the commenter details described above. Meta leaves a separate direct-message connection behind, which you remove under Connections. Google leaves the access you granted on your account, which goes when you close it. It does not empty your inbox: to remove everything, email us to close the account. Facebook and Instagram users have a dedicated path described on the data deletion page. Once we can tell which records are yours, we complete a deletion request within 30 days. Closing a Rumore account also clears the activity records tied to it.

A few things do not follow that path, and we would rather list them than let you assume otherwise.

Our records of work the system ran for you, like sending a message, sometimes carry the phone number or email address it went to. We delete those about a month after that work finishes.
We cannot pick out the scorecard visit records described above at all, because we hold no IP address or browser in them to match you against. They expire on their own within a year.
Ask, where you put questions to your AI Employee in your own words, keeps what you typed as the record of that conversation. When a customer asks to be erased we delete every exchange in which Ask looked that person up, question and answer together, but a name you only mentioned in passing leaves us nothing to match on. Those messages go when your retention window closes, which is a year unless we have agreed otherwise with you.
When your AI Employee writes something for you, a price list, a flyer, a note for your staff, we keep that document as well, along with the sentence you asked for it in. Erasing a customer removes the documents from any exchange in which Ask looked that person up, and as above, a name you only typed leaves us nothing to match on. One thing works differently here and we would rather say it. A document’s retention window runs from the last time you edited it rather than the day it was written, so a document you keep correcting stays with us for as long as you keep correcting it. You can delete any document yourself, from your documents list or from the conversation it came from, and its text is removed from both straight away. The rest of that conversation goes on the window described above rather than with the document: the sentence you typed to ask for it, the document’s title, and your AI Employee’s reply, which can describe what the document said.
When you tell your AI Employee to remember something about your business and you approve it, we keep that sentence and read it into its later answers for that Location. Unlike the conversation and the documents above, it has no retention window: it stays until you delete it from the list in Settings (on the billing page while your subscription is off) or close your account. Deleting it removes it straight away, together with its text on the card where you approved it. The conversation where you told it goes on the window described above rather than with the memory, so for a few more questions your AI Employee can still see what you said there. Erasing a customer removes a remembered sentence only when it was saved in an exchange where Ask looked that person up; a name you only typed leaves us nothing to match on.
When you ask your AI Employee to do something on a schedule and you approve it, we keep that request and its schedule until you delete it from the list in Settings (on the billing page while your subscription is off) or close your account. We run it as the person who set it up, or as whoever last resumed it, until you pause or delete it or that person no longer has access to the Location. While your subscription is off, it does not run. Like a remembered sentence, it has no retention window. What each run writes goes into your conversation, on the same window as the rest of it. Erasing a customer deletes a routine that was set up in an exchange where Ask looked that person up; a name you only typed leaves us nothing to match on.
When you attach a photo or a file for your AI Employee to read, we keep the file itself for a day and then delete it. What your AI Employee said about it stays in your conversation, on the same window as the rest of it.
When you speak to your AI Employee instead of typing, your browser turns your voice into text through a speech service that it chooses, not us: in Chrome on Android and on a computer that is Google, on an iPhone and in Safari it is Apple, and in Edge it is Microsoft. We never receive the recording, only the words you then choose to send.
Your AI Employee can read its answers aloud. Reading aloud stays off until you turn it on. Your browser does the reading, not us, in the best voice it has. Some of those voices come from the browser maker’s own speech service: in Chrome that is Google, and in Edge it is Microsoft. When one of those voices reads, your browser sends that answer’s text to that company to be spoken. Voices on an iPhone and in Safari are made on the device.
If you joined the waitlist, including through the scorecard before it stopped asking, that signup keeps your email address separately, and closing an account does not remove it. Write to privacy@rumore.ai and we will delete it.
We keep a single line recording that the deletion happened, with nothing personal in it, because we have to be able to show that we did it.

Changes to this policy

If we change this policy in a way that matters, we will update the date at the top and, for significant changes, tell account holders by email before the change takes effect.

Contact

Inevitable AI
privacy@rumore.ai