Privacy Policy

01Who we are

Warble is a service operated by UseWarble, LLC, a Florida limited liability company, at 124 N Nova Road #118, Ormond Beach, FL 32174. In this policy "we", "us" and "Warble" mean that company. The service is provided through usewarble.com and the Warble app.

We decide what happens to the information described here, which makes us responsible for it. If you want to ask us anything about it, or make any of the requests in section 13, the address is hello@usewarble.com. It is a real inbox and it is the only address you need.

Warble is in early access and is sold to businesses rather than to consumers. You must be 18 or over to hold an account.

02A United States service

Warble is intended for businesses in the United States. It is not offered to customers in the United Kingdom or the European Union, it is not designed around their rules, and we do not hold ourselves out as meeting them. Our servers, our storage and almost all of the providers listed in section 9 are in the United States.

About US state privacy laws, honestly

California's Consumer Privacy Act (as amended by the CPRA) and the similar laws now in force in other states apply to companies above set thresholds — annual revenue, or the number of consumers whose information they handle, or revenue earned from selling personal information. Warble is a small company and does not currently meet those thresholds. So we are not going to put a badge on this page claiming compliance with a law that does not yet apply to us. If that changes, this section changes with it.

What we do instead, as a matter of policy rather than because a statute forces us to:

• Anyone can ask for a copy of the personal information we hold about them.
• Anyone can ask us to correct it.
• Anyone can ask for an export of their account's data.
• Anyone can ask us to delete it.

We offer those to everybody, in every state, because they are the right things to offer — not because we have worked out who is entitled to them. Section 13 says how to ask.

03What we collect, and why

Account identity

Sign-in, sessions and organisation membership are handled by Clerk, a third-party authentication provider. Clerk holds your sign-in credentials; Warble never sees or stores your password. From Clerk we read your email address, your first and last name, your Clerk user ID, and which organisations you belong to and in what role. In our own database we store your Clerk user and organisation IDs, the organisation name, and a cached copy of the organisation owner's email address, used to route notifications.

Why: to give you an account, keep other people out of it, scope every record to the right organisation, and contact you about the service.

Content you and the AI create

Ideas, post bodies, per-channel rewrites, translations, campaigns, schedules, publishing status and any error a platform returned. Anything you type into the composer is content too — a brief, an edit, or a post you wrote yourself. You own it; see the Terms of Service.

Usage, credits and cost records

An append-only ledger of credit grants and debits with a reason for each, and one row per AI generation attempt recording which surface requested it, which provider and model ran, how many credits were charged, what it cost us, whether it succeeded, any error, how long it took, and the result of the automated quality check.

Why: to meter usage fairly, refund failed generations, and understand our own costs.

Email we send you

Warble sends transactional email — for example when a customer conversation needs a human. Those records store the recipient address and whether the email sent, was skipped or failed. We do not send marketing email today. If we ever start, it will be something you can decline, and this page will say so first.

Abuse controls

Public endpoints are rate-limited using the requesting IP address. That counter lives in the server's memory, is never written to the database, and is lost when the service restarts.

04The website scan

Onboarding asks for one thing: your website address. We then fetch your homepage and up to three further pages on the same domain, read the text, and hand it to an AI model to produce a business profile.

We store the URL you gave us, the list of pages we actually read, and what the model derived from them: a summary of the business, the industry, who the audience is, the offer and call-to-action found on the site, four to six content pillars, any proof points on the site (reviews, statistics, milestones), and a brand kit — colours, fonts, logo and image direction, photography direction, ad-creative style, tagline, voice adjectives and a differentiation line.

We only follow links on the domain you gave us, we only fetch over HTTPS, and we refuse addresses that resolve to internal or private networks. If your site blocks a direct read we may retry through a fetching proxy, which receives the URL being analysed and nothing else (section 9).

Why: this profile is what makes the generated posts sound like your business rather than generic marketing. Without it the product does not work.

A scan also creates a session record holding a token, the site URL, the derived brand and profile, and the first posts generated for you, so that the preview you saw before signing up becomes the account you then created.

05Images and video you upload

You can upload your own photos and video. Files go to object storage and are served from a CDN URL; the database row keeps that URL, the original filename, the content type, the pixel dimensions, tags, a SHA-256 content hash and the file size. Before your first upload we record a one-time confirmation, per workspace, that you have the rights to the media you upload, and we store that timestamp.

Screening

Automated screening may be applied to images you upload. Where it is switched on, the image is sent to Google Cloud Vision's SafeSearch classifier before it is stored, and an image the classifier marks as adult or violent content is refused. That path is built into the software but is not active today — it only runs once we configure it, and section 9 lists it as inactive.

Please read that as what it is. Screening is a general classifier, not a guarantee of anything: it misses things, it is not a check on copyright, likeness or consent, and it is not a substitute for your own judgement about what you upload. Assume nothing you upload is checked, and do not upload anything you do not have the rights to or anything featuring a person who has not agreed to appear.

06Connected accounts and what we hold

You can connect Instagram, Facebook, LinkedIn, X, TikTok, Threads, YouTube, Pinterest, Bluesky and Google Business. The connection is made through our social publishing provider, not by Warble directly: you are sent to a single-use authorisation page hosted by that provider and you grant access there.

Disconnecting is immediate. When you disconnect a channel, we call the provider to revoke the connection first and then delete our record of it — the identifiers are gone at that point, not queued for later. You can also revoke Warble's access from inside the social platform itself, and doing both is the belt-and-braces version.

Nothing publishes unattended. Warble generates drafts; a post only goes out because someone on your side scheduled it.

If you run accounts for a client — an agency, a freelancer, a marketing team — that is allowed, and by connecting the account you are confirming you have the client's permission. You remain responsible for what is published. See the Terms of Service.

07What comes back from your channels

Post performance

For posts published through Warble we periodically pull back per-post metrics — views, likes, comments, impressions, reach, saves, shares, clicks — and keep them as a time series so you can see how a post moved. We also take a daily snapshot of each connected channel's follower count, and we keep that history ourselves because the publishing provider only retains about four weeks of it.

Tracked links

When a post carries a Warble link, we mint a short /c/<key> redirect. Each hit is logged with the link key, the post it belongs to, the browser's user-agent string, the referring URL, a timestamp, and a flag for whether it looks automated. We do not store the visitor's IP address in that log. The IP is used only in memory, briefly, to rate-limit abuse.

Comments and direct messages

If you turn on the inbox, comments and DMs on your connected channels flow into Warble. For each conversation we store the platform, the thread identifier, and the other participant's handle, display name and platform ID; for each message, the direction, the text, any attached media, the platform's message ID and timestamps. Replies you send are stored the same way, and a conversation that cannot be handled is escalated to you in the app and can be emailed to you.

That material belongs to people who never dealt with us. We hold it only so you can read and answer it. We do not use it to generate content for anyone else, we do not build profiles from it, and we do not sell it. If one of those people contacts us about a message sitting in your inbox, we will pass the request to you rather than act on your account's data without you.

08What we send to AI model providers

Warble generates content by calling AI models run by other companies. This is not incidental — it is how the product works, and it means your brand information leaves our servers. Here is what goes where.

What is never sent to any of them: your password (we never have it), your connected-channel authorisations, your credit ledger, or billing details.

One case deserves a specific mention. The deeper website analysis that runs after you sign in uses a model with a server-side web search tool enabled, capped at two searches per analysis. The model may therefore issue live search queries derived from your industry and business context, and those queries reach the provider's search infrastructure.

The posts Warble produces are generated by these models. They can be wrong and they can invent things. The compliance screening feature does not change that — it flags risky language, it is not legal advice and it is not a guarantee of compliance. See the Terms of Service.

09Subprocessors — who else receives data

These are the external services Warble actually calls. Each row says what the service does and what specifically reaches it. Services we do not use are not listed, and nothing has been included for completeness.

Each of these has its own terms and its own handling of data. We will keep this list current, and where a change of provider materially affects you we will say so under section 16.

10What we do not do

• We do not sell personal information, and we do not share it with advertising networks or data brokers.
• We do not train our own models on your content, and we do not permit training on it where a provider offers that control — see section 8.
• There is no third-party web analytics, advertising pixel or session-recording tool on the Warble site. We went looking in our own codebase because we wanted to be able to say this: there is none.
• We do not store your passwords — not your social account passwords, and not your Warble one, which Clerk holds.
• We do not log the IP addresses of people who click your tracked links.
• We do not read one customer's workspace to make content for another.
• We hold no security certification and make no compliance claim. We are not SOC 2 audited and not ISO 27001 certified, and section 2 says where we stand on state privacy laws. A badge we have not earned is worse than no badge.

11How long we keep things

The short version: while your account is open, we keep what the product needs to work. When you close it, we delete.

• Your account, workspace, brand profile, website scan, generated content and media are kept for as long as the account exists, and are deleted within 30 days of the account being closed.
• Connected-channel identifiers are deleted immediately when you disconnect a channel — see section 6.
• Generated media that never gets used is cleaned up on its own: an image or video generated for you but not attached to a scheduled post is swept and deleted about a day later. Media that is scheduled or published is kept.
• Post performance readings are a time series, throttled so a live post adds only a small number of rows per day, and follower snapshots are one row per channel per day. Both are kept while the account is open, because the source only retains about four weeks.
• Click logs and the credit ledger are append-only and are kept while the account is open.
• Inbox conversations and messages are kept while the account is open. You can delete a conversation sooner, and you should if a customer asks you to.
• Backups. Deleting something from the live system is not the same instant as deleting it from every backup copy. Where backups exist, deleted data ages out with them rather than disappearing the moment you ask.

12Security — what actually exists

What follows is a description, not a promise. No system is perfectly secure, and we hold no certification that says otherwise.

What is in place

• All traffic is served over HTTPS, with certificates issued and renewed automatically.
• Authentication is delegated to Clerk. We never handle your password. Every route other than the public marketing pages, the tracked-link redirect, the signed webhook and the health check requires a signed-in user.
• Every record is scoped to an organisation or workspace, and queries are written to enforce that boundary. The incoming webhook that delivers comments and DMs works out which workspace an event belongs to from our own records, and never trusts a workspace identifier supplied in the payload.
• That webhook is rejected outright unless its signature verifies against a shared secret, before any work is done.
• Fetching your website is guarded against server-side request forgery: HTTPS only, and loopback, private and link-local addresses are refused — including hosts that resolve to one, and on every redirect hop.
• The tracked-link redirect only ever forwards to an http or https destination, so a stored value cannot turn our domain into a phishing hop.
• Public endpoints are rate-limited.
• Media identifiers are unguessable random UUIDs, and uploaded bytes are content-hashed.

What is not in place, and you should know it

• Media URLs are public to anyone holding the link. They are not access-controlled.
• Automated screening of uploaded images is built but not switched on. Treat uploads as unchecked.
• Rate limiting is per-server and held in memory; it resets when the service restarts.
• We are a small team. There is no 24-hour security operations centre, and we do not pretend there is.

If something does go wrong and your information is affected, we will tell you what happened, what we know, and what we are doing about it — by email to the address on your account, without waiting to be asked.

13Your choices, and how to use them

Email hello@usewarble.com. That is the whole process. Tell us what you want and we will do it — we may need to check you are who you say you are first, and we will come back to you within 30 days.

• A copy of the personal information we hold about you.
• A correction to anything that is wrong.
• An export of your account's data.
• Deletion of your account and its data.
• To stop a particular use you object to — tell us which, and we will either stop or explain plainly why we cannot.

What we can and cannot do

• We can export or delete your account, workspaces, brand profile, generated content, media, inbox history, analytics and click logs.
• We cannot delete posts already published to your social channels. Those live on the platform under your account — delete them there.
• We cannot retrieve data on your behalf from an AI model provider. What a provider retains has to be asked of that provider.
• Deletion lags in backups, as section 11 explains.

Showing your posts as examples

Our Terms of Service give us a limited right to show posts generated on Warble as examples of what the product does — the generated post and the brand look, never your uploaded photographs of identifiable people. If you would rather we did not, email hello@usewarble.com and say so. There is no form and no argument; we will stop.

Marketing email

We do not send any today. If that ever changes, every message will carry a one-click unsubscribe, and this page will be updated before the first one goes out.

14Cookies and the marketing site

Warble sets very little, and none of it for advertising.

There is no analytics cookie, no advertising cookie and no tracking pixel on our site, which is why you are not being shown a consent banner. Our pages do load a typeface from Google Fonts, which means your browser contacts Google and Google sees your IP address; see section 9. If we ever add an analytics tool, this section and section 9 will be updated before it goes live.

15Children

Warble is a business tool sold to businesses. It is not directed at children, it is not designed for them, and you must be at least 18 to hold an account.

We do not knowingly collect personal information from children. If you believe a child's information has reached us — including through a comment or DM arriving in an inbox — tell us at hello@usewarble.com and we will delete it.

16Changes to this policy

We will update this page when what we do changes — most likely when billing is switched on, when image screening is switched on, or when a provider in section 9 changes. The "last updated" date at the top always reflects the current version.

Where a change materially affects you, we will email the address on your account before it takes effect, rather than quietly editing the page and hoping you look.

17Contact

One address for everything — questions about this policy, requests under section 13, and ordinary support:

• Email — hello@usewarble.com
• Post — UseWarble, LLC, 124 N Nova Road #118, Ormond Beach, FL 32174

For how the service itself works — what you may publish, who owns what, and how credits and billing will work — see the Terms of Service.