Guardian — Privacy Policy

Guardian (guardian.alwayshere.app) is a product of AlwaysHere LLC.

Effective date: August 2, 2026
This policy lives at: guardian.alwayshere.app/guardian-privacy
Contact: privacy@alwayshere.app
Mailing address: 584 Hyacinth Place, Highland Park, IL 60035

This policy covers Guardian only. AlwaysHere’s other products have their own policy at alwayshere.app, because they handle data differently. Guardian’s terms of service are at guardian.alwayshere.app/guardian-terms.

The short version

You are probably tired. Here is the whole thing in thirty seconds.

• Guardian is free. There is no card, no trial, no paid tier. We have written no payment code at all.
• There is one account, and it is yours. The adult Guardian supports does not have a login. They have no page, no password, no way to sign in.
• That means you are giving us information about someone who never signed anything. Their name, their age, their diagnosis in your words, what upsets them, what they love. That is the most important sentence in this document.
• We ask you to tell them. Guardian is built to be told about, not hidden. If they don’t know it’s there, this product is not doing what it was built to do.
• Guardian cannot see anything except what is typed to it. Not Facebook, WhatsApp, Instagram, Snapchat, Discord, Twitch, or Tinder. Not their texts, their email, their browser, their screen, or their location. Not even their other AI chats, if Guardian isn’t loaded in them.
• Guardian cannot watch. It has no clock and no scheduler. It runs only in the instant their AI calls it, and it stops the moment that call is answered. It cannot check on anyone, notice a quiet week, or detect that someone stopped talking.
• Everything Guardian records, it records because the person chose to say it in a chat Guardian was loaded into.
• Guardian does not store their conversations. It stores short summaries a model wrote — capped at 600 characters — and only when it decides a moment was worth writing down.
• Their conversations go to Anthropic, because Guardian runs inside the Claude account they already use. That account is theirs, not ours. Anthropic’s terms apply to it.
• When Guardian texts you about a danger sign, that text goes through Twilio and your phone carrier. It contains a short summary of a bad moment, with their first name in it.
• Can a human at AlwaysHere read this data? Yes. We hold the key to our own database. Anyone who says otherwise about their own product is lying to you. What we promise instead: Guardian holds no credentials to anything your family owns — no email password, no inbox token, no key to any account of yours. It cannot log in anywhere on its own. And every read and write it does is written to a log you can read back.
• We do not sell data and we do not run ads. There is no ad code and no analytics vendor in this product.
• There is no delete button yet. Email privacy@alwayshere.app and we delete it by hand. We are saying that plainly instead of pretending a button exists.
• This is not HIPAA-covered, not medical care, and not an emergency service. If someone is in danger, call 911. For crisis support, call or text 988.

The rest of this page is the same thing, in detail.

1. Who we are

AlwaysHere LLC operates Guardian at guardian.alwayshere.app. We are a United States company, organized in Minnesota. We decide what data Guardian collects and how it is used, so under US privacy law we are the party responsible for it.

Guardian is free and always will be. We say that in the policy, not just the marketing, because it is the honest explanation for a question you should be asking: if I’m not paying, what’s the catch? The answer is that there is no data business behind this. We do not sell, rent, trade, or share personal information for money or for anyone else’s advertising. There is no advertising code, no data broker, and no analytics vendor in this product. If that ever changes, this policy changes first, and we will email you before it takes effect.

2. Two people, one account. Why that matters.

Guardian involves two people and treats them very differently. You need to understand this before you set it up.

The account holder — usually a parent, family member, or a paid caregiver. You create the account. You answer the setup questions. You get the dashboard, the journal, the alert log, and the text messages. Everything Guardian stores is attached to your account.

The supported adult — the person Guardian is for. They have no account and no login. There is no page they can sign into, no password, no settings screen. Their entire relationship with Guardian happens inside their own Claude conversations, through Guardian’s tools.

So: we hold personal information about a person who is not our customer, did not create the account, and never clicked “I agree.” Some of that information is health-adjacent — a birthdate, a diagnosis or condition in your words, notes about distress and self-harm risk.

We are not going to bury that. It is the central fact of this product, and everything below follows from it.

3. Consent — please read this part

By finishing setup, you represent to us that at least one of the following is true:

1. The supported adult knows Guardian is there and has agreed to it. You have told them, in words they understand, that an AI layer is in their chats, that it keeps a shared journal, and that it will text you if certain things come up. They said yes.
2. You hold legal authority to make this decision for them — you are their guardian, conservator, or hold a power of attorney or an equivalent that covers this, and you are acting inside that authority and in their interest.

If neither is true, please do not set Guardian up.

We want to be honest about how strong that representation currently is: we do not verify it, and today we do not record it. There is no checkbox, no attestation stored in our database, no document upload. It is your word, and you are the one who lives with it. We are working on capturing it properly.

We also want to be honest about the harder problem underneath it. Capacity is not a switch. Many of the adults Guardian is built for can absolutely understand and agree to it — and are the best judges of what they want in their own chats. Some cannot, or can on some days and not others. A legal guardianship does not automatically mean a person has no view about what happens in their own conversations. Our position, plainly stated:

• Tell them. Guardian is designed to be known about. It is built to say what it is when asked, to say plainly that it shares safety summaries with you, and never to claim it keeps private notes.
• Guardian is removable. If the supported adult wants it gone from their Claude, they can remove the connector themselves. We consider that a feature and not a defect. But note what removal does and does not do — see Section 12.
• A safety layer someone resents catches nothing. Consent is not just the legal answer here. It is the thing that makes the product work at all. Guardian only ever learns what the person decides to tell it, so a person who does not want it there tells it nothing.

A limit you should know about. Guardian’s instructions tell it to be transparent — to answer honestly if the supported adult asks what gets shared, and never to pretend it keeps secret notes. Those are instructions to a language model. They are not a guarantee about what the model says in any given conversation. The parts we actually enforce on our servers are described in Section 6.

4. What Guardian cannot see

This section is near the top on purpose. Most privacy policies describe what a product collects. The more important fact about Guardian is how narrow the opening is in the first place.

Guardian has no visibility into any of this:

• Facebook, WhatsApp, Instagram, Snapchat, Discord, Twitch, Tinder, or any other app or social network
• Text messages and email
• The browser, browsing history, or anything on the screen
• Location, at any level of precision
• Contacts, calendar, photos, files, microphone, or camera
• Other AI conversations, including other Claude conversations, where Guardian has not been loaded

Guardian is not an app on anyone’s phone and it is not installed on any device. It is a connector inside a single AI conversation. It cannot open an app, read a thread, or look at anything it was not handed inside that conversation.

There is no scheduler. This is worth stating precisely, because people reasonably assume a safety product is watching. It is not. Guardian’s server runs code only when an inbound request arrives from the AI mid-conversation, and it stops as soon as that request is answered. There is no background job, no timer, no polling, and no process that runs while nobody is typing. In practical terms:

• Guardian cannot check on anyone.
• Guardian cannot notice silence, absence, or a person who has stopped talking.
• Guardian cannot detect a change over time, a missed routine, or a bad week — except by reading what the person types to it in a new conversation.
• Guardian cannot contact anyone first. Every text you receive is downstream of something that was said to Guardian moments earlier.

So: everything Guardian records, it records because the person chose to say it, in a chat Guardian was loaded into. There is no other input. That is the whole surface area, and it is the reason the setup questions matter more than any alert threshold — a safety layer someone does not enjoy talking to hears nothing at all.

5. What we collect, and exactly where each piece comes from

5.1 From you, when you create an account

• Email address and password, or a Google sign-in, if you choose that instead. Accounts are handled by Supabase Auth. If you use Google, Google learns that you signed into Guardian.

5.2 From you, during setup (the setup questions)

Everything in this list is typed by you into the setup form and stored in our guardian_profiles table:

About the supported adult

• Their name, and what they like to be called
• Their birthdate
• Their condition or diagnosis, in your words (the form’s own example: “Autistic, with some anxiety”)
• Which language they prefer about themselves (identity-first, person-first, or “ask them”)
• Topics they love
• Topics to never bring up
• Routines that matter to them
• How they like to be spoken to: sentence length, energy, emoji on or off
• Whether Guardian should check in during a conversation, and how often
• What Guardian should do when they say no
• The exact calming phrases that work in a dark moment
• Optional history notes — “anything from the past Guardian should quietly know”

About you

• Your name, and what they call you (“Mom”, “Grandpa Joe”)
• Your phone number for alert texts, and optionally a second number
• Your alert threshold (how sensitive Guardian should be)
• What Guardian may share with you and what should stay private
• Whether you want alert texts at all, and whether you want email summaries later

A note on that second phone number. If you enter one, alert texts go to both. That means a person who never used this product, never agreed to anything, and may not know Guardian exists will receive automated messages containing your loved one’s name and a summary of a hard moment. Please ask them first, and please tell them where the messages come from. You can remove the second number from your setup at any time.

How birthdate and condition are actually used. Your birthdate entry never leaves our servers as a date — Guardian’s instructions receive only a computed age (“Danny is 27”). The condition text is passed into Guardian’s instructions, sanitized and capped at 200 characters, labelled as calibration only, with a hard instruction never to raise it as a topic. Both shape how Guardian speaks. Neither is meant to become subject matter.

Things that never reach the AI at all. Your phone numbers, the install key, your alert threshold, and the history notes are marked server-only. They are deliberately excluded from the instructions Guardian receives, so they cannot be read out of a chat or talked out of the model.

5.3 Generated for you

• An install key — a random 24-character value that identifies your Guardian. It is embedded in the connector URL you paste into Claude. Anyone holding that URL can reach your Guardian, which is why the install page tells you to keep it private.

5.4 From Guardian, while it runs

Remember Section 4: every one of these exists only because something was typed to Guardian in a live conversation.

• Journal entries (guardian_journal) — a short plain-language summary written by the AI, capped at 600 characters, plus an optional one-word mood and a safety reading of “clear”, “watch”, or “concern”. These are written when the model decides a moment was worth recording. They are not transcripts. We do not store your loved one’s conversations.
• Shared memory (guardian_memory) — durable facts, each a short label plus up to 500 characters: people in their life, routines, preferences, boundaries, health basics, goals, interests. Capped at 200 facts; when it fills, the oldest unpinned fact is deleted.
• Safety alerts (guardian_alerts) — the category, a summary capped at 400 characters, a confidence level, whether it was flagged imminent, whether a text was actually sent, any send error, and which number(s) it went to.
• A tool log (guardian_tool_log) — one row every single time Guardian uses one of its tools: which tool, when, and a short note (for example, Saved memory "sister Mia" or financial / high — caregiver texted). This is the log we keep pointing at. It exists so that neither we nor Guardian can quietly do something you cannot see.

5.5 From the website

• If you use the “get in touch” form for an organization: your email, optional name, and a one-way hash of your IP address — we store the hash, never the raw IP, so we can stop bots without keeping a record of where you were.

5.6 What we do NOT collect

• No conversation transcripts.
• No location data. No device fingerprinting.
• No contacts, calendar, photos, or files.
• No advertising or analytics tracking in this product.
• Nothing from their other apps, texts, DMs, or any chat Guardian is not part of. See Section 4 — Guardian has no ability to look outward, and no ability to run when nobody is talking to it.

6. What Guardian’s tools actually do

Guardian has exactly eight tools. This is the complete list of everything it can read or write. Each one runs only when the AI calls it during a live conversation.

Every one of those calls writes a row to the tool log. There is no ninth tool, no hidden endpoint, and nothing that runs on a timer.

Things the server decides, not the AI. This matters for your trust in the alerts, so here is exactly how it works. The AI can report a concern. It cannot decide whether you get a text. Our server does, using rules the AI never sees:

• Your threshold setting gates it. If you turned texts off entirely, the alert is still recorded — it just never leaves as a text.
• No more than 1 text per category per 6 hours.
• No more than 4 texts per day.
• Never two texts closer than 15 minutes apart.
• A near-identical concern inside 30 minutes does not re-text you.
• A hard ceiling of 6 texts per 12 hours that nothing overrides — because past that point a phone is being flooded and more texts stop helping. The alerts are still recorded; only the texting stops.
• The one exception: an alert flagged as imminent self-harm bypasses the daily cap.

What the AI is allowed to put in a text. Only the summary. Everything around it is a fixed template written by us. Before sending, the summary is stripped of links and anything shaped like a phone number, and capped at 240 characters — so an alert text can never become a delivery route for something a bad actor typed into a chat.

The five settings the supported adult can change from inside their chat, and nothing else: what to call them, talk more or less, emoji on or off, energy level, and adding a topic to never bring up. They cannot change safety rules, alerting, or thresholds from inside a conversation, and an attempt that reads like a rule change is refused.

7. Who else receives this data

These are every company that touches Guardian data, what each one gets, and why.

Anthropic (Claude) — the big one. Guardian runs inside the supported adult’s own Claude account. That account belongs to them (or to you), not to us. The entire conversation goes to Anthropic, exactly as it would if Guardian did not exist, because they are already talking to Claude. Guardian adds its instructions and its tool calls to that conversation. Anthropic’s own privacy policy and terms govern that account and that conversation — we have no control over it and cannot delete anything from it. If you want to know how conversation data is handled at Claude, read Anthropic’s policy for the account you are using.

Lovable and Supabase — hosting, the database, and authentication. Everything described in Section 5 lives in a Postgres database operated by Supabase through Lovable’s cloud platform. Our server code uses an administrative database key that bypasses the per-user access rules. That key is what makes the whole product work: Guardian is reached by an anonymous request from Claude, so it cannot be authenticated as you. It also means our server code — and any person at AlwaysHere holding that key — can technically read and write every row. See Section 9.

Twilio, and your mobile carrier — alert texts. When an alert clears the threshold and the rate limits, we send one HTTPS request to Twilio containing your phone number and the message body. Twilio hands it to your carrier. The message contains: the word GUARDIAN, the alert category, the supported adult’s first or preferred name, and the sanitized summary of the distress moment. Text messages are not encrypted end to end. Twilio can see the body. Your carrier can see the body. Anyone holding your unlocked phone can see it on the lock screen. That is an unavoidable property of SMS, and it is the reason the summary is short and the reason we sanitize it.

Lovable’s managed email service — transactional email to you, the account holder. Two messages exist today: a “here’s how to get back in” message when you save partial setup, and a setup-complete message. The setup-complete email contains your install key and connector URL. Treat that email the way you’d treat a password reset email.

Google — only if you choose “Continue with Google” to sign in. Then Google knows you have a Guardian account, and Google’s own privacy policy governs the sign-in. If you sign up with an email and password instead, Google is not involved at all.

That is the complete list. There is no analytics vendor, no advertising network, no CRM, and no data broker.

We may also disclose data if we are legally required to — a valid subpoena, court order, or law enforcement demand — or where we believe in good faith that disclosure is necessary to prevent imminent physical harm. If we are legally permitted to tell you about a demand for your data, we will.

If AlwaysHere LLC is ever sold or merged, data may transfer to the acquirer. We will tell you before that happens, and the terms of this policy travel with the data.

8. What we do not do

Only claims we can actually back:

• We do not sell your data, and we do not share it for anyone’s advertising. No advertising or analytics code exists in this product.
• We do not train AI models on your data. We do not build models. We send text to Anthropic’s API and get text back; anything beyond that is governed by Anthropic’s own terms for the Claude account being used.
• We do not store conversations. The tools that write to our database write summaries, and the caps — 600, 500, 400 characters — are enforced by the database itself, not just by the code.
• We do not monitor anyone. There is no scheduler, no background process, and no way for Guardian to observe a person who is not actively typing to it. See Section 4.
• We do not send Guardian’s instructions your phone numbers, install key, history notes, or alert threshold. They are excluded on purpose so they cannot be extracted from a chat.
• We do not let the AI decide when to text you. See Section 6.
• Guardian does not hold credentials to anything your family owns. No email password, no inbox token, no key to any of your accounts. It cannot log in anywhere on its own. It sees what is handed to it inside a conversation, and nothing else.

9. “But could a human at AlwaysHere read this?”

Yes.

We hold the administrative key to our own database. Any vendor who tells you they physically cannot see data sitting in a database they operate is either confused or lying to you, and one screenshot would disprove it.

So here is the real commitment, which is a commitment about behavior rather than a claim about capability:

• We access individual family data per incident, with a reason — you asked us for support, a bug needs diagnosing, or we are legally required to. Not routinely, and not out of curiosity.
• We do not read journals for product research, marketing, or “quality review.”
• Everything Guardian itself does is logged in the tool log, in your account, where you can read it back. If Guardian read your memory, wrote a journal entry, or made a safety call, there is a row for it.

We will tell you honestly where that record stops: the tool log covers Guardian’s actions, not ours. There is not yet a customer-facing log of AlwaysHere staff database access, and we are not going to pretend there is. If that matters to you — and it reasonably might — email privacy@alwayshere.app and ask what has been accessed on your account. We will answer.

Don’t trust the promise. Trust the log, and hold us to the part the log does not yet cover.

10. How long we keep things

Straight answer: right now, indefinitely, until you ask us to delete it.

There is no automatic expiry on journal entries, alerts, or the tool log. The only automatic deletion in the entire system is the shared memory cap — once there are more than 200 facts, the oldest unpinned ones are removed to make room.

We consider open-ended retention a problem, not a feature, and we intend to set real limits. Until we do, this section says what is true rather than what sounds good. When we set retention periods, this policy will say what they are, and we will email account holders before they take effect.

Waitlist and organization-contact entries are kept until you ask us to remove them.

11. Deleting your data

There is no delete button in the app today. We are telling you that instead of describing a self-service flow that does not exist.

To delete everything, email privacy@alwayshere.app from the address on the account. Say you want your Guardian deleted. We will:

1. Confirm it is you, using the account email.
2. Delete the account and, with it, the profile, every journal entry, every alert, every memory fact, and the tool log. These are linked to your account with a database rule that removes them together, so this is a real deletion and not a hidden flag.
3. Confirm to you in writing when it is done. We aim to complete deletions within 30 days.

What we cannot delete for you:

• The Claude conversations themselves. Those live in the Claude account, which belongs to you or to the supported adult. Delete them there. Anthropic’s policy governs how long they keep them.
• Alert text messages already delivered. They are on your phone and in Twilio’s and your carrier’s delivery records.
• Emails already delivered to your inbox.
• Backups, for a short window. Deleted rows can persist in our hosting provider’s routine backups until those rotate out.

You can delete individual memory facts yourself, right now, from your dashboard. Guardian can also delete one from inside a chat when the supported adult asks it to forget something. Journal entries and alerts cannot currently be individually deleted, by you or by us through the app — ask us and we will do it directly.

To stop Guardian without deleting anything: remove the connector from Claude, or turn off text alerts in your setup. Removing the connector stops all new data immediately — and because Guardian only ever runs when it is called, “immediately” is literal. There is nothing left running in the background.

12. The supported adult’s rights, even though they have no account

They did not sign up. It is still their information. These rights are ours to give, not the law’s to require, and we are giving them.

The supported adult, or anyone acting for them, can:

• Ask what we hold about them. Email privacy@alwayshere.app. We will produce a plain-language copy of the profile, the memory, the journal, and the alert history.
• Ask us to correct something that is wrong.
• Ask us to delete it. See below for how we handle a conflict.
• Ask Guardian to forget a fact, right now, in conversation — Guardian has a tool for exactly that, and it takes effect immediately.
• Remove Guardian themselves. Deleting the connector in their own Claude account stops Guardian from running. Nothing we build overrides that; it is their AI account.

Be honest about what they can see today. From inside a conversation, the supported adult can ask Guardian what it remembers about them — the whole shared memory list, up to 200 facts — and Guardian can bring back the last 10 journal entries. They cannot see the alert list from inside a chat, and there is no page for them to log into. If they want the full picture, including alerts, they can ask their caregiver, or they can email us at the address above and we will send it to them.

Be clear about what removing the connector does not do. It stops new data. It does not delete the data already stored, and it does not revoke the account holder’s install key. Only a deletion request does that.

When the two people disagree. If the supported adult asks us to delete their data and the account holder objects, we will not just take the account holder’s side because they hold the login. We will:

1. Acknowledge the request directly to the person who made it.
2. Contact the account holder.
3. Where there is no legal guardianship in place, treat the supported adult’s own wishes about their own information as decisive.
4. Where there is a guardianship, weigh it — and still, at minimum, stop Guardian from running for that person while we work it out.

We are not pretending this is a solved problem. It is a genuinely hard one and we would rather write down how we will approach it than leave it silent.

Identity checks. We will make sure a request is genuine before acting on it, particularly a deletion. We will not require documents that a person may not have; we will ask questions we can check against the account.

State privacy laws. Some US states give residents rights to access, correct, delete, and opt out of sale or targeted advertising. We do not sell data or run targeted advertising at all, so there is nothing to opt out of. For the rest, use the email above and we will honor the request regardless of which state you live in. We will not discriminate against anyone for exercising a right.

13. Children

Guardian is for adults. Do not set it up for anyone under 18.

We are naming a real gap rather than hiding it: nothing in the product currently stops you. We ask for a birthdate, but we do not block a birthdate under 18, and we do not verify anyone’s age. That is a gap we intend to close.

Until we do, this is on the record: setting Guardian up for a minor is against our terms. If we learn a Guardian profile describes someone under 18, we will contact the account holder, and we will delete the profile and everything attached to it. If you believe a minor’s information is in Guardian, email privacy@alwayshere.app and we will remove it.

Guardian is not directed to children and we do not knowingly collect information from anyone under 13.

14. Security

Specifically, and without overselling it.

What we do:

• Everything travels over HTTPS.
• Every table has row-level security on, restricting reads to the account that owns the row.
• The administrative database key is server-side only. It is never sent to a browser.
• The install key is 96 bits of randomness, unique per family, and never echoed back in any error response.
• A Guardian install key does not work until setup is finished — a half-finished setup has no verified phone number, which means no way to raise an alarm, so we refuse the key.
• Everything typed into setup is sanitized before it reaches the AI’s instructions, so nobody can inject fake instructions that outrank the safety rules.
• Alert text bodies are stripped of links and phone-number-shaped text before sending.
• Every tool call is logged.

What we want you to know honestly:

• The install key is a password in a URL. Anyone who has that URL can reach your Guardian: read the journal, read the memory, and write to both. There is currently no way to rotate or revoke it from the app. If you think it has leaked, email privacy@alwayshere.app and we will replace it by hand.
• Alert texts are not encrypted end to end. See Section 7.
• We do not offer two-factor authentication on Guardian accounts yet.
• We are not HIPAA-covered. AlwaysHere LLC is not a covered entity or a business associate, and Guardian is not a HIPAA-compliant service. We collect information that is health-adjacent — a condition in your words, notes about distress — and we protect it seriously, but we do not claim a legal standard that does not apply to us. If you are a provider or an organization with HIPAA obligations of your own, Guardian in its current form is not built for that, and you should not put patient information into it.
• No system is unbreachable. If a breach affects your data, we will tell you promptly and tell you what we know, including what we do not yet know.

15. The honest limits

This belongs in the privacy policy because these limits shape what the data means.

• Guardian only sees the conversations it is part of. It cannot see other apps, texts, DMs, social media, or any chat it is not in. Everything it catches, it catches because the supported adult chose to tell it.
• Guardian cannot watch, and cannot notice absence. There is no scheduler. If your loved one stops talking to Guardian, Guardian does not know and cannot tell you. A quiet journal means Guardian was not spoken to. It does not mean everything is fine.
• Guardian is not guaranteed monitoring. It is an extra layer of support. No AI catches everything. It will miss things.
• Guardian is not an emergency service. It cannot call 911. It cannot dispatch anyone. If someone is in danger right now, call 911. For crisis support in the US, call or text 988.
• Guardian is not therapy, medical care, legal advice, or financial advice, and it is built to refuse to give directives in any of those areas.
• A journal entry is a language model’s summary, not a record of fact. It reflects what the model understood. It can be wrong, incomplete, or miss the point. Do not treat it as evidence.
• Guardian can be removed by the person using it. We consider that correct. If the journal goes quiet, that may be why.

16. Where we operate

AlwaysHere LLC is a United States company. Guardian’s servers and database are operated in the United States by our hosting providers, and this policy is written for US law.

We are not going to claim a restriction we do not enforce: nothing in the product currently blocks a signup from outside the United States, and the phone number field accepts international numbers. If you are outside the US, understand that your information will be stored and processed in the United States, under US law, and that we do not currently offer the specific mechanisms that GDPR, UK GDPR, or other non-US regimes require. If you are in the EU, the UK, or another such jurisdiction and you want your data removed, email privacy@alwayshere.app and we will delete it.

17. Governing law and disputes

This policy is governed by the laws of the State of Minnesota, without regard to its conflict-of-laws rules.

Disputes about this policy or about how we handle your data are resolved the same way as every other dispute with us: by binding individual arbitration, as set out in the Guardian Terms of Service. Those terms include the arbitration agreement, the class-action waiver, the opt-out window, and the exceptions — read them there rather than here, so there is one version of that language and not two.

Nothing in this section stops you from contacting a government regulator, a state attorney general, or a privacy authority. We would rather you email us first, but you do not have to.

18. Changes to this policy

If we change something that matters — a new company receiving data, a new category of data collected, a new use — we will email account holders before the change takes effect and update the date at the top. Small clarifications and fixes get the date update alone.

We will not quietly widen what we do with your information. If we ever want to do something with this data that you would not expect from reading this page, we will ask you first.

19. Contact

Email: privacy@alwayshere.app
Post: AlwaysHere LLC, 584 Hyacinth Place, Highland Park, IL 60035

Write to us about: getting a copy of your data, correcting it, deleting it, a request from or on behalf of the supported adult, a leaked install key, a question about what a human at AlwaysHere has accessed, or anything in this policy you think is not true.

That last one especially. If something on this page does not match what Guardian actually does, tell us and we will fix the page or fix the product.

Guardian is a product of AlwaysHere LLC. Free, always. If someone is in immediate danger, call 911. For crisis support in the US, call or text 988.