HappyLike โ† the sky

Privacy Policy

Last updated: 2026-11-13

This policy explains what HappyLike, LLC ("we," "us," "our") collects when you use HappyLike, why, who else sees it, how long we keep it, and what you can make us do about it. It is incorporated into our Terms of Service.

We have tried to write it as a description of how the product actually works rather than as a list of everything the law would let us do. Where the product does something that could surprise you โ€” sending your posts to an automated classifier, reading the words inside photographs โ€” it is described plainly below rather than covered by a general phrase.

1. What we collect

What you give us. An email address, and a password if you choose one. A username, a first and last name, and a display name. Anything you choose to add to your profile: a bio, a location, links, a profile picture, a cover image. If you buy advertising, your business name, address, phone number and industry.

What you make here. Posts, replies, the photographs attached to them, reactions, saved posts, who you follow, who is in your community, who you have blocked or muted, reports you file, appeals you make, and messages you send us through the contact form.

What the service records as you use it. When your account was created and when its details changed. Your IP address โ€” when your account is created, when you sign in, when you ask for a sign-in link, when a credential on your account changes (a password, an email address, a second factor), and when you upload a photograph. Two reasons, and we would rather name both: so that "somebody changed my password" and "somebody got into my account" have answers, and because the law that governs reports of child sexual abuse material lets a report include the address material was uploaded from, which is the single most useful thing in it. Section 7 says what we send and to whom. Which browser session asked for a sign-in link, as a one-way hash that never leaves our systems. Whether a message we sent you bounced or was marked as spam, so we stop sending to an address that cannot receive.

We keep the addresses from sign-in links for ninety days, and the rest for as long as we keep your account. None of them is used to advertise to you or to build a profile of you, and we do not use them to work out where you are.

What automated screening produces. A category, a confidence figure, a positivity score and a short written reason for each post and reply; whether an image was refused; the text found inside images; and, for a picture, a fingerprint of the file itself and whether it matched something we have already removed. See section 4.

We do not use advertising or analytics trackers, build profiles of you across other websites, buy data about you from anybody, or ask for your contacts, your precise location, or access to your device beyond what you actively use.

2. Why, and on what basis

We use the above to run the service you asked for: to show you the sky, to deliver replies and notices, to let people find each other by handle, and to keep your account working. Where the UK and EU GDPR applies, that is performance of a contract.

We use it to keep the place safe: to screen content, to act on reports, to apply the published Community Guidelines and our Acceptable Use Policy, to detect abuse, and to rate-limit requests. That is our legitimate interest in running a service people are not harmed by, and in some cases a legal obligation.

We use it to bill advertisers and keep records of what was bought. That is a contract with them, and a legal obligation to keep the records.

We use your name and email address to write to you about HappyLike itself โ€” that we have opened, what has changed, and occasional news about the place. We would rather be plain about this than bury it: creating an account puts you on that list, we did not ask you to tick a box for it, and every message carries a one-click unsubscribe link that works immediately and for good. Leaving the list costs you nothing else โ€” your account is untouched and you go on getting the messages the service has to send you, like a sign-in link or a notice about something you reported. Where the UK and EU GDPR applies this is our legitimate interest in telling the people who joined something about the thing they joined, and you can object at any time by unsubscribing or by writing to us. It is not advertising: nobody else writes to you through it, nobody buys access to it, and we do not sell or rent the list.

Where we ask for consent we will say so and you can withdraw it. We do not rely on consent for anything described here.

3. What other people see

Your posts, replies, reactions and profile are public. Anybody, signed in or not, can read them, and search engines may index a profile page. Treat everything you write as something a stranger will read, because it is.

These are not public: your email address, your phone number, your real name (we hold your first and last name and publish neither โ€” your display name is whatever you chose), reports you file, who you have blocked or muted, whether you asked for a lighter sky or for strong language to be masked, your saved posts, and how many people follow you.

Your community is private in both directions. Nobody can list it, and every connection in it was agreed to by both people.

Reactions name the people who made them to the person whose post it is, and to anybody who opens that post. There are no counts anywhere on the surface.

4. Automated screening, and what it does not decide

Every post and reply is sent to Anthropic's Claude API to be classified. Images are analysed by Amazon Rekognition, which includes reading any text in them โ€” people put words in pictures to get past text screening, and a filter that cannot see them is not a filter.

A picture is also compared against the files we have already removed. Every upload is held somewhere private until it has been checked, and part of checking it is measuring it: an exact digest of the file, and a short "perceptual" fingerprint that survives a picture being resized or re-saved. Those are compared against the fingerprints of material this platform has previously taken down for child safety or as a non-consensual intimate image. It is a comparison of files rather than a reading of content โ€” it can tell us a picture is one we have seen before, and it can tell us nothing at all about one we have not. The comparison happens on our own systems and the list is ours; the picture is not sent anywhere for it.

What that produces is a category, a score, a reason, occasionally a flag that somebody may be at risk, and for a picture a fingerprint and whether it matched.

What it can do on its own: withhold clearly violating content from publication, hide a reply until a person looks at it, rank a post lower, and put something in front of a human reviewer.

What it cannot do on its own, with a single exception: suspend or remove an account. That is a person's decision everywhere but one place. If a picture matches a file already identified as child sexual abuse material, the account is closed and the report is filed without a person having looked first. We do it that way because the match identifies a file rather than forming a judgement about a person, and because the law (18 U.S.C. ยง 2258A) requires the report as soon as we actually know โ€” and the match is knowing.

When a decision is made about your content you are told, with a reason, and you can contest it โ€” see section 8 and /appeals. If you are in the UK or EU, Article 22 gives you the right not to be subject to a purely automated decision with a significant effect on you. The one automated account decision described above is made to comply with a legal obligation, which is a basis that Article allows, and an appeal against it is read by a person โ€” that human review is the safeguard the right asks for, and it is the reason the exception is narrow enough to write down in one sentence.

Content sent for classification is not used to train anybody's models. It is sent to be read and scored, and the answer is stored with the post.

5. Who else handles it

We keep the list short, and each of these is a processor acting on our instructions:

Amazon Web Services Hosting, database, file storage, image resizing, image screening (Rekognition), and email delivery (SES). United States.
Anthropic Classifying the text of posts and replies. United States.
Cloudflare Sits in front of the site, filters attack traffic, and sets a cookie to tell people from bots.
Stripe Advertiser payments only. Card details go to Stripe and never reach us.
Sentry Error reports when something breaks. Configured not to send personal data.

The mailing list is ours, not a marketing platform's. The software that holds it runs on a server we control, so your name and address are not handed to a third-party newsletter company, and nothing in it is shared with anybody. The messages themselves go out through the email provider in the table above.

We do not sell personal information, and we do not share it for cross-context behavioural advertising.

Advertising here can be aimed at a place, and never at a person. An advertiser may ask for their advertising to be shown in a country, a state or province, or a US metro area. We work out which of those you are in from your connection, on the request that draws the page, and that is the whole of it:

  • Nothing is written down. Not the address, not the place, not that you were shown anything. The page is drawn and the answer is forgotten.
  • We never use the state you set in your settings for this. You told us that so we could show you a support line you could actually dial, and using it to choose an advertisement would be putting it to a purpose you did not agree to. The two are kept apart in the code on purpose.
  • The advertiser is told nothing about where you are. They see how many people were shown their advertising and how many tapped it. There is no breakdown by place, and there is no column anywhere that could produce one.
  • Nothing about you is used to choose it โ€” not what you post, not what you read, not what the automated screening thought of either. The only thing about you that changes which advertising you see is that you see none at all for an hour after you post something difficult, and that is a protection rather than a target.

Because your connection is what is read, a VPN or a holiday will show you advertising for somewhere you are not. That is the trade, and it is deliberately the direction that costs nothing.

See the Advertiser Agreement.

We will disclose information if the law actually requires it. If we are allowed to tell you, we will โ€” with one exception we would rather state than have you discover: we do not tell somebody that we have reported them for content involving a child. Telling them is how evidence disappears, and that one is not ours to choose.

We also report such content to the National Center for Missing & Exploited Children, which is a disclosure the law requires rather than a company processing data for us, which is why it is not in the table above.

If the business is ever sold or merged, information may transfer with it, and you will be told before anything about this policy changes for you.

6. Cookies

Three, and none of them advertise to you:

  • A session cookie, so you stay signed in. It cannot be read by JavaScript.
  • A CSRF token, so a form on another site cannot act as you.
  • Cloudflare's bot-management cookie.

There is nothing here to opt out of, because there is no tracking to switch off.

7. How long we keep it

Your account, when you delete it: 30 days, then it is destroyed. Deleting is immediate from everybody else's point of view โ€” your posts, replies and pictures leave the service at once โ€” and the 30 days exist so an account deleted at midnight in anger can be brought back. After that a nightly job destroys the account, its content and its files, and it cannot be recovered.

Your username is held for a year after that, and cannot be taken by anybody else in the meantime, because it appears in other people's old posts and in links people have already sent.

Moderation records outlive the account, deliberately. What was decided, when, and why, is kept so that repeat behaviour stays countable and so an appeal can be judged later. Once the account is destroyed those records identify it only by a one-way keyed hash, which cannot be reversed into an account and holds nothing about a person.

The record of changes to your account โ€” who changed a password, a handle, a role, and when โ€” is destroyed with the account, because it holds your old handles and addresses. The one case where part of it is kept instead is the preservation exception at the end of this section.

Sign-in links last 15 minutes and work once. Sessions last two hours.

You stay on the mailing list until you leave it or your account is destroyed. Unsubscribing takes you off it at once. Deleting your account takes you off it too โ€” when the nightly job destroys the account it removes your address from the list in the same pass, so nothing keeps writing to somebody who has gone.

Sent mail: we keep the fact that an address bounced or complained, so we stop sending to it.

One exception, and it is the only one. If we have reported content to the National Center for Missing & Exploited Children under 18 U.S.C. ยง 2258A โ€” which we do for anything sexual involving a child, and nothing else โ€” the law makes that report a request to preserve the material for a year, and we keep a copy for that long in a place built so that nothing, including us, can delete it early.

It covers more than the picture, because the statute says it does. Alongside the file we keep a record of where it came from: the email address on the account, the username and display name, when the account was registered, the IP addresses of its most recent sign-ins and credential changes, the address the file itself was uploaded from, and the post the file was attached to. All of that is copied at the moment the report is raised, which is what makes it outlive the account being destroyed. It is the only place in this policy where something we promise to delete with your account is kept anyway, and we would rather write it down than have you discover it.

And that record is sent, not only kept. The report to the National Center for Missing & Exploited Children carries the account's address, its username and display name, a link to its profile, an identifier of ours that means nothing outside our own systems, the IP addresses above, and โ€” if we have suspended the account permanently โ€” that we have done so and when. The law allows all of it and leaves the choice to us, and we make it deliberately: a report that says only "this file was on our platform" is one nobody can act on. It goes to that organisation and, through them, to law enforcement, and to nobody else. We do not tell the account holder that a report was filed, for the reason given in section 5.

Everything else about the account is still destroyed on the ordinary schedule above, and a deletion request is still honoured in full for all of it.

Separately, the fingerprint of removed material is kept indefinitely. Not the picture โ€” the measurement of it described in section 4, which is what stops the same file being uploaded here again and holds nothing that identifies anybody. See section 9.

8. What you can make us do

See it and take it with you. Settings โ†’ Download my data gives you a file containing your account, your posts and replies, your reactions, and the moderation decisions made about your content. Other people's data is left out: a report filed about you identifies whoever filed it, and that is theirs rather than yours.

Correct it. Your profile is editable in Settings. A username can be changed once every 30 days.

Delete it. Settings โ†’ Delete my account, and then the 30 days above. You do not have to give a reason and we will not ask for one.

Contest a decision. /appeals shows every moderation decision about your content and lets you contest it. An appeal is reviewed by a different person than the one who made the original decision.

Object, restrict, or withdraw. Write to us and we will do it, or explain why we cannot. For the mailing list described in section 2 you do not have to write to anybody: the unsubscribe link at the foot of every message is the whole of it.

We answer within 30 days. We do not charge for any of this, and we will not make the service worse for you because you asked.

Use the contact form โ€” it works whether or not you can sign in, which is deliberate, because the people who most need it are often locked out. If you are in the UK or EU you may also complain to your data protection authority.

9. Children

You must be at least 18. We do not knowingly collect anything from a child, and if we find that we have, we delete it. If you believe an account belongs to somebody under 18, tell us through the contact form โ€” it works whether or not you have an account.

This section used to say 13, which was wrong: the product has never admitted anybody under 18, and the sign-up form refuses a date of birth that does not clear it.

Content involving a child is the one thing we do more than delete. Anything sexual involving somebody under 18 is reported to the National Center for Missing & Exploited Children, which is who US law (18 U.S.C. ยง 2258A) requires us to report it to and who passes it to law enforcement. The report can include the material itself, the account it came from and the technical details of the upload. We keep that material โ€” and a record of the account and the upload behind it โ€” for at least a year afterwards because the law requires us to, which is the one exception to the deletion promises in section 7, and which that section sets out in full.

10. Where your data is

The service runs in the United States and the third parties above are United States companies. If you are in the UK or EU, your information is transferred there โ€” we rely on the standard contractual clauses our providers offer for that transfer.

11. Keeping it safe

Everything travels over HTTPS. Passwords are hashed and never stored in a form we can read. Two-factor secrets are encrypted. Sign-in links are stored hashed, expire, and work once. Access to the moderation tools requires a second factor. We are a small team and we do not pretend to be immune: if something happens that affects you, we will tell you rather than wait to be asked.

12. Changes

If we change this policy in a way that materially affects you, we will tell you at least 14 days beforehand, the same notice the Terms promise. The date at the top always says when it last changed.

13. Contact

HappyLike, LLC St. George, Utah, United States Contact us

All legal documents