Guardian — Terms of Service

Guardian is at guardian.alwayshere.app. These Terms are a contract between you and AlwaysHere LLC (“AlwaysHere,” “we,” “us”). They cover Guardian only. AlwaysHere’s other products have their own terms, because Guardian carries risks they do not.

Section 19 requires most disputes to be resolved by binding individual arbitration and waives your right to a jury trial and to participate in a class action. You can opt out of arbitration within 30 days. Read Section 19.

The short version

What Guardian is. Guardian is a free layer you add to the Claude account your loved one already uses. It teaches that chat how they communicate, keeps a plain-language journal, and can send you a text message when it spots a danger sign inside that chat.

What Guardian is not. It is not an emergency service. It is not a monitoring service. It is not medical or clinical care. No human at AlwaysHere is watching your loved one’s chats or reviewing alerts.

What Guardian cannot see. Guardian sees one conversation: the chat it is loaded into. It cannot see Facebook, WhatsApp, Instagram, Snapchat, Discord, Twitch, Tinder, text messages, email, their browser, their screen, their location, or their AI chats where Guardian is not loaded. Everything Guardian knows, it knows because your loved one chose to type it. And it only runs while they are typing to it — it is not watching in the background, and it cannot notice that they have gone quiet.

Three things to understand before you use it:

1. Guardian will never call 911. It does not contact police, paramedics, or a crisis line for anyone. If someone is in danger right now, call 911. For a mental health crisis, call or text 988. Guardian is not a substitute for either.
2. A text may not arrive. Alerts are filtered by the sensitivity setting you chose and by rate limits that stop your phone being flooded. On top of that, the AI has to notice and report the concern in the first place, and a text message has to survive your carrier and your phone. Any one of these can fail. Guardian is an extra layer. It is never the only layer.
3. The person Guardian is for has to know about it. Guardian is set up on one account — yours. Your loved one does not sign anything. By using Guardian you are promising us that they know it is there and agreed to it, or that you legally hold the authority to agree for them. Watching an adult in secret is against these Terms, and depending on where you live it may be against the law.

If any of that is not acceptable to you, do not use Guardian. That is a real option and we would rather you take it than rely on something that will not do what you need.

1. Accepting these Terms, and what the words mean

1.1 How you accept

You accept these Terms by ticking the acceptance box when you create your Guardian account, whether you sign up with an email address or with Google. That tick is a signature. We record the version of the Terms you accepted and the date and time you accepted it, and you can ask us for that record at any time.

If you keep using Guardian after we change these Terms, you accept the change under Section 13.

If you do not accept these Terms, do not create an account. If you have one, delete it under Section 14.

1.2 Definitions

• “Guardian” or “the Service” — the Guardian website, the caregiver account and dashboard, the Guardian connector (an MCP server) that runs inside a third-party AI chat, and the text and email messages we send.
• “You” or “Caregiver” — the person who creates the account, completes setup, and holds the account. Guardian has exactly one account per household setup, and it is yours.
• “Supported Person” — the adult Guardian is set up for. They do not have a Guardian account.
• “Claude” — the AI assistant made by Anthropic, PBC. Anthropic is not affiliated with AlwaysHere and does not endorse Guardian.

2. Guardian is not an emergency service

Read this section before anything else.

Guardian does not call 911. It does not contact police, fire, emergency medical services, a mobile crisis team, a suicide hotline, a school, a doctor, a case manager, or an adult protective services agency. It has no ability to do any of those things. There is no code in Guardian that places a call to anyone.

No human at AlwaysHere monitors Guardian. Nobody reads the journal in real time. Nobody reviews alerts. Nobody is on call. If Guardian records a serious concern at 3 a.m., the only person who may learn about it is you, by text, and only if every condition in Section 3 is met.

In an emergency, call 911. If someone is in immediate danger of harming themselves or being harmed, call 911.

For a mental health or suicide crisis, call or text 988 (the 988 Suicide & Crisis Lifeline, in the United States). Guardian will share the 988 number inside the chat when it detects self-harm content, but sharing a phone number is not the same as making a call. Only a person can make the call.

Do not use Guardian as an alternative to any of this. If your loved one’s safety plan currently depends on a human being reachable, keep that human. Guardian is added on top of that plan. It does not replace any part of it.

3. Guardian is not guaranteed monitoring

Guardian is an extra layer of support. It is not a safety net, and we will not describe it as one.

Guardian can miss things. Over enough time, it will miss things. That is not a defect we intend to fix; it is what the product is.

3.1 Guardian only sees the conversation it is part of

Guardian is a connector inside one AI chat. That chat is the whole of its world.

Guardian has no visibility into any of the following: Facebook, WhatsApp, Instagram, Snapchat, Discord, Twitch, Tinder or any other app; text messages; email; phone calls; the web browser; the screen; the device’s location; and any other AI assistant, or any Claude conversation, where Guardian is not loaded. It has no camera, no microphone, and no location awareness. It does not read anything on the device.

Everything Guardian knows, it knows because your loved one chose to type it into a chat Guardian was loaded into. If a scammer reaches them on a game chat, Guardian learns about it only if they mention it to Guardian. If they never mention it, Guardian never knows, and neither will you.

3.2 Guardian is not running when nobody is typing to it

This is the single most important limitation in these Terms, and the one caregivers most often misunderstand. Please read it twice.

There is no scheduler. There is no background process. Nothing in Guardian runs on a timer. Guardian executes only when your loved one’s AI chat sends it a request — which happens only while they are actively typing to it. In between those moments, Guardian is not running. It is not watching, waiting, listening, or checking.

The consequences follow directly:

• Guardian cannot detect silence or absence. If your loved one stops chatting — for an hour, a day, a month — Guardian does not notice, because nothing is running to notice with. It will not text you to say they have gone quiet. It cannot.
• Guardian cannot notice a change over time on its own. It does not review its own journal on a schedule, does not trend anything, and does not look back for patterns unless it is asked inside a live conversation.
• Guardian cannot check on anyone. It never initiates contact. It cannot start a conversation, send a message, or reach out to your loved one for any reason.

Silence in the journal means nobody is chatting. It does not mean everything is fine. Do not read an empty dashboard as reassurance. It carries no information about your loved one’s safety at all.

3.3 An alert can fail to reach you, for any of these reasons

1. The AI did not report it. Guardian’s safety check has to be called by the AI model in your loved one’s own account. If the model does not recognise a danger sign, or judges it differently than you would, no alert is created. We do not control that model (see Section 5).
2. Your sensitivity setting filtered it. You chose one of three thresholds during setup. On “Balanced” a low-confidence concern is recorded but does not text you. On “Only when Guardian is confident,” only high-confidence concerns text you.
3. You turned texting off. If text alerts are switched off in your settings, alerts are recorded in your dashboard and never leave as a text.
4. Rate limits suppressed it. Guardian enforces server-side limits so that a caregiver’s phone is not flooded: a minimum gap between any two texts, a cooldown per concern category, a daily cap, and an absolute ceiling of six texts in a rolling twelve-hour window. A concern flagged as imminent self-harm overrides the duplicate window, the minimum gap, the daily cap and the category cooldown — those limits will not hold it back. It does not override the twelve-hour ceiling. Once six texts have gone out in twelve hours, texting stops until the window clears, no matter how serious the next alert is. Alerts are still recorded in your dashboard during that period; they simply do not leave as a text.
5. A near-duplicate was suppressed. If a very similar concern in the same category was already texted recently, the repeat is not sent — unless it is flagged as imminent self-harm.
6. The text failed to deliver. We hand messages to a telecommunications provider (Twilio). We do not control whether your carrier delivers them, whether they are filtered as spam, or whether your phone is on, charged, or in signal. We do not retry failed sends, and we do not confirm delivery to your handset. A message accepted by the carrier is not a message you read.
7. Something was down. Our servers, our database host, our SMS provider, or Anthropic can all fail independently of each other.

Every safety check Guardian runs is recorded in your dashboard whether or not a text went out, along with the reason no text was sent. That record is the honest account of what happened. Do not treat the absence of a text as evidence that nothing happened.

3.4 What we do not promise

We do not promise that Guardian will detect any particular risk, that it will detect risk within any particular time, that any alert will be delivered, that the Service will be available at any given moment, or that the journal is a complete record of anything.

4. Guardian is not medical, legal, psychological, or financial advice

Guardian is not a clinician, therapist, counsellor, physician, nurse, social worker, case manager, care provider, lawyer, or financial adviser. It is software.

Using Guardian does not create a doctor-patient, therapist-client, attorney-client, fiduciary, or professional care relationship of any kind, between you and AlwaysHere or between the Supported Person and AlwaysHere. AlwaysHere is not a covered entity, a business associate, a care provider, or a mandated reporter, and Guardian is not a medical device, a clinical decision support tool, or a diagnostic instrument. It is not designed, tested, or certified for any of those purposes.

Guardian is instructed never to give medical, legal, or financial directives. It is an AI, so it may do so anyway. If it does, that output is not advice from AlwaysHere and you should not act on it.

Nothing Guardian records is a clinical record. The journal is a plain-language summary written by an AI. It is not a chart, not an assessment, and not evidence of anything.

5. The AI is not ours, and AI output is unpredictable

Guardian runs inside the Supported Person’s own Claude account, which is Anthropic’s product, governed by Anthropic’s terms, and controlled by whoever holds that account. We do not own, host, run, or control that account, that model, or its outputs. Anthropic may change, degrade, price, restrict, or discontinue Claude or its connector support at any time, and that would break Guardian without any action by us.

What we actually control is the tool layer on our own servers: the system instructions Guardian supplies, the validation of everything the model sends us, the alert thresholds, the rate limits, and the wording of the text message itself. The model fills in one summary field; our server writes the rest of the message and strips links and phone numbers out of it before sending. Server-side enforcement is meaningfully stronger than asking a model nicely, and it is why threshold and alerting rules cannot be talked out of Guardian from inside a chat.

It is still not a guarantee. Guardian’s conversational rules — the red lines about romance, secrecy, honesty, and not pulling someone away from real people — are delivered to the model as instructions. Instructions are followed most of the time, not all of the time.

The model can produce output that is wrong, confusing, offensive, upsetting, or harmful, including to a vulnerable person, and including in direct contradiction of the rules we give it. You accept that risk when you install Guardian. If you are not willing to accept it, do not install Guardian.

6. Consent and authority — your promise to us

This is the most important promise you make in these Terms, and it is the one we will enforce.

Guardian runs on a single account: yours. You complete setup, you write the personalization answers, you set the alert threshold, and you supply the phone number. The Supported Person signs nothing and agrees to nothing in writing. Their protection is this section.

By creating a Guardian account and completing setup, you represent and warrant that at least one of the following is true, and that it remains true for as long as you use Guardian:

• (a) The Supported Person knows Guardian is installed, knows it keeps a journal, knows it can text you about safety concerns, and has agreed to it; or
• (b) You hold current legal authority to consent on their behalf — guardianship, conservatorship, a healthcare or durable power of attorney, supported decision-making authority, or an equivalent under the law where you both live — and you are acting within that authority.

You further agree that:

• You will not install or keep Guardian on an adult’s AI account covertly. Guardian is designed to be told about. Its instructions require it to answer honestly when asked what it shares and with whom, and the safety topics it always reports are shown to you during setup so that you can show them to your loved one. Defeating that design is a breach of these Terms.
• If the Supported Person tells you, or tells Guardian, that they want it removed, you will remove it. Guardian is deliberately easy to remove. That is respect, not a flaw.
• You will not use Guardian to monitor a spouse, partner, employee, tenant, roommate, client, or any other adult who has not consented and over whom you hold no legal authority.
• The Supported Person is not aware of, and Guardian does not create, any obligation to keep using it.

Covertly monitoring an adult without their consent or your legal authority may be illegal under state wiretapping, electronic surveillance, stalking, and elder-abuse and vulnerable-adult statutes, and the consequences can be criminal. We are not your lawyer and this is not legal advice about your situation. If you are unsure whether you have authority, get advice before you install Guardian, not after.

If we learn or reasonably believe you have breached this section, we may terminate your account immediately and without notice.

7. Eligibility

• You must be 18 or older to create a Guardian account.
• The Supported Person must be 18 or older. Guardian is built for autistic adults and adults with intellectual and developmental disabilities. Its language, its assumptions about autonomy, and its safety design are all built for adults.
• Guardian is not for minors. Do not set it up for anyone under 18.

We currently ask for the Supported Person’s date of birth during setup to help Guardian calibrate how it speaks, and we do not use it to block anyone. That means the age limit above depends on you being truthful. If you enrol a person under 18, you are in breach of these Terms, and we may suspend or delete the account and its data on becoming aware of it. If you believe a minor has been enrolled, email legal@alwayshere.app and we will delete the profile and its associated journal, memory, and alert records.

8. Your account and your install key

You are responsible for your login credentials and for everything done under your account.

Setup produces a connector URL containing a personal install key. That key is the only credential protecting your Guardian data. Anyone who has that URL can read the journal and the shared memory for your Supported Person, add or delete memories, and cause Guardian to behave as if it were installed in their chat. Treat it like a password: do not post it, do not email it to people who do not need it, do not put it in a shared document.

If you believe the key has been exposed, email legal@alwayshere.app and we will issue a new one. We are not responsible for access obtained through a key you disclosed.

9. What Guardian stores, and who it goes to

We are specific about this because vague privacy language is how products hide things.

What you give us at setup: your name and email, what your loved one is called, their date of birth, a free-text description of their condition or diagnosis in your words, their interests, topics to avoid, routines, communication preferences, calming phrases you wrote, optional background notes, your alert sensitivity setting, and one or two phone numbers for text alerts.

What Guardian writes while it runs: plain-language journal entries, a list of durable facts about the person (“shared memory”), a record of every safety check including ones that did not text you, and a log of every tool call Guardian made.

Where that data goes:

• Anthropic, PBC — the conversation itself happens in Claude. Everything Guardian sends into the chat, and everything said in it, is processed by Anthropic under Anthropic’s terms and privacy policy, not ours. Guardian’s setup instructions and the recalled journal entries are sent into that chat.
• Lovable and Supabase — our application hosting and our database. Guardian’s records live there.
• Twilio and mobile carriers — text alerts. The message body contains what your loved one is called, the alert category, and a short summary of the concern. Once it leaves us, it is handled by Twilio and by your carrier under their terms.
• Lovable’s managed email service — account and setup emails to you.

What the Supported Person can see today: Guardian is instructed to answer honestly, in the chat, about what it shares and with whom, and the shared memory and recent journal entries are read back into the chat by the recall and memory_list tools, so the Supported Person can ask Guardian directly and get a true answer. There is currently no separate login for the Supported Person. The dashboard, with the full journal and alert history, is visible to the account holder only. We consider this a gap and intend to close it; until we do, do not tell your loved one they have their own account, because they do not.

We do not sell your data, and we do not use Guardian journal, memory, or alert content to train AI models.

You should also read the Guardian privacy policy at guardian.alwayshere.app/guardian-privacy for the full detail. Where these Terms and that privacy policy conflict about data, the privacy policy controls.

10. Text messages

By entering a phone number during setup and leaving text alerts on, you consent to receive automated safety text messages from Guardian at that number.

• Enter your own number. We do not verify that the number you enter belongs to you. If you mistype it, a text describing a private safety concern about your loved one — potentially including self-harm — will be delivered to a stranger. Check it carefully. Check it again if you change it.
• Only enter numbers you are entitled to give us. If you add a second number, you confirm the person who holds it has agreed to receive these messages.
• Message and data rates may apply. Frequency varies with what Guardian encounters and is capped by the rate limits in Section 3.3.
• You can turn text alerts off at any time in your dashboard. Alerts continue to be recorded there.
• To ask about message delivery, email legal@alwayshere.app.

11. Acceptable use

Do not:

• Use Guardian to monitor anyone in breach of Section 6.
• Use Guardian for anyone under 18.
• Attempt to make Guardian conceal itself, lie about what it shares, or deny that it reports safety concerns.
• Attempt to disable, bypass, or manipulate the safety thresholds, rate limits, or red lines, including through instructions written into setup fields.
• Enter another person’s personal information into setup fields without a lawful basis to do so.
• Share, sell, or publish install keys, or attempt to access any profile that is not yours.
• Attack, scrape, overload, reverse engineer, or probe the Service or its infrastructure, or use it to send unlawful, harassing, or abusive messages.
• Resell Guardian, or present it to third parties as a monitoring, compliance, or duty-of-care solution — for example, to satisfy a supervision requirement at a care organisation. It is not one, and doing so puts a vulnerable person at risk on the strength of a claim we have never made.

12. Guardian is free, and we can change or end it

Guardian is provided free of charge. There is no fee, no subscription, no trial, and no payment mechanism in the product. We are not charging you today and we have no plan to charge existing caregivers for what Guardian does today.

Being free is not a promise of permanence. We may change, limit, suspend, or discontinue Guardian, in whole or in part, at any time. Guardian also depends on third parties — Anthropic, Twilio, our hosting providers — any of whom can change or end their service in ways that break ours. Where we discontinue Guardian deliberately and it is practical to do so, we will give reasonable advance notice to the email on your account so you can adjust your loved one’s safety plan and export what you want to keep.

Because the Service is free, you have paid nothing for it, and nothing in these Terms should be read as a paid support commitment, a service level agreement, or an uptime guarantee. There are none.

13. Changes to these Terms

We may update these Terms. When we make a material change — particularly to Sections 2 through 6, or to Sections 14 through 17 — we will post the updated Terms with a new effective date and email the address on your account at least 14 days before they take effect. Continuing to use Guardian after the effective date means you accept the change. If you do not accept it, delete your account before then.

14. Termination, and what happens to your data

You can stop at any time. Remove the Guardian connector from the Claude account, and Guardian stops functioning immediately — it has no independent existence outside that chat.

We can stop at any time. We may suspend or terminate your account, with or without notice, if you breach these Terms — especially Sections 6, 7, or 11 — or if we discontinue the Service.

Your data on termination. To delete your account and everything associated with it, email legal@alwayshere.app from the address on your account. We will delete your profile, journal, shared memory, alert history, and tool log, and revoke the install key. We aim to complete deletion within 30 days. Some data may persist for a limited period in routine backups and in third-party systems we do not control — notably the conversation history inside the Supported Person’s Anthropic account, which only they or the account holder can delete, and delivery records held by our SMS and email providers under their own retention policies. We currently have no automatic retention or deletion schedule; records persist until you ask us to remove them.

Sections 4, 5, 9, 15, 16, 17, 18, and 19 survive termination.

15. Disclaimer of warranties

Guardian is provided “as is” and “as available,” without warranties of any kind, express or implied. To the fullest extent permitted by law, we disclaim the implied warranties of merchantability, fitness for a particular purpose, title, and non-infringement.

Specifically, and without limiting the above, we make no warranty that:

• Guardian will detect any danger, risk, scam, exploitation, self-harm indication, or crisis;
• any alert will be generated, will be generated in time, or will be delivered to you;
• the journal or shared memory is accurate, complete, or a faithful account of what was said;
• the AI’s output will be appropriate, safe, correct, or consistent with the instructions we give it;
• the Service will be uninterrupted, timely, secure, or error-free.

Some jurisdictions do not allow the exclusion of certain implied warranties, so parts of this section may not apply to you.

16. Limitation of liability

Read this section. It is the part of these Terms that a court is most likely to be asked about.

Guardian is free. That means the usual software cap — “our liability is limited to what you paid us” — would limit our liability to zero, which is not a fair thing to write and not something we want to hide behind. So we are stating our position plainly instead.

What we do not accept liability for. To the fullest extent permitted by law, AlwaysHere is not liable for any harm arising from an alert that was not generated, not sent, delayed, suppressed, or not delivered, or from anything the AI said or failed to say. That includes self-harm, suicide, injury, death, financial loss, fraud, exploitation, abuse, emotional distress, loss of privacy, or loss of data. This applies whether the claim is in contract, tort (including negligence), strict liability, or any other theory, and whether or not we were told such harm was possible.

This is not a technicality. It is the direct consequence of Sections 2 and 3: Guardian is not an emergency service and does not guarantee monitoring, so it cannot be relied upon as one, and liability cannot be built on reliance we have told you not to place.

Total cap. To the fullest extent permitted by law, AlwaysHere’s total aggregate liability to you for all claims relating to Guardian will not exceed one hundred U.S. dollars (US$100). We state a real number rather than “the amount you paid,” because you paid nothing and a zero cap is not an honest term.

What we do not disclaim. Nothing in these Terms excludes or limits liability that cannot be excluded or limited by law — including liability for our own fraud, fraudulent misrepresentation, gross negligence, or wilful misconduct, or for death or personal injury caused by our negligence, where applicable law prohibits that exclusion. Some jurisdictions do not allow the exclusion or limitation of incidental or consequential damages, so parts of this section may not apply to you. In that case, our liability is limited to the smallest amount the law allows.

This allocation of risk is the basis of the bargain. Guardian is offered free to families who need it. It is offered on the understanding that it is an extra layer, not a guarantee, and it could not be offered at all on any other terms.

17. Indemnification

You agree to indemnify, defend, and hold harmless AlwaysHere LLC and its members, officers, employees, and contractors from any claim, demand, loss, liability, damage, or expense (including reasonable legal fees) arising from:

• your breach of these Terms, in particular your representations about consent and authority in Section 6 and about age in Section 7;
• a claim by the Supported Person, or by anyone acting on their behalf, that Guardian was installed or operated without their knowledge, consent, or lawful authority;
• personal information you entered about another person without a lawful basis;
• your use of Guardian in a professional, organisational, or care-provider setting, or any representation you made to a third party about what Guardian does; or
• your use or misuse of the Service generally.

We may take over the defence of any matter subject to this section at our own expense, and you agree to cooperate.

18. Governing law

These Terms are governed by the laws of the State of Minnesota, United States, without regard to its conflict-of-laws rules, and by applicable United States federal law. The arbitration agreement in Section 19 is governed by the Federal Arbitration Act.

Nothing in this section deprives you of the protection of any mandatory consumer-protection law of the state where you live that cannot be waived by agreement.

Guardian is offered in the United States. If you use it from outside the United States, you do so on your own initiative and are responsible for compliance with your local law — including consent and surveillance law, which is materially stricter in many countries than in the United States.

19. Dispute resolution — arbitration and class-action waiver

Please read this section carefully. It affects your legal rights. It requires most disputes to be resolved by an individual arbitrator instead of a judge or jury, and it prevents you from bringing or joining a class action. You can opt out, and Section 19.3 tells you exactly how.

19.1 Talk to us first

Before starting arbitration or any other formal proceeding, email legal@alwayshere.app with a description of the dispute, what you want, and your account email. Give us 30 days to try to resolve it. We will do the same before bringing a claim against you. Most problems are faster to fix than to litigate, and this step costs you nothing.

19.2 Binding individual arbitration

If we cannot resolve it, you and AlwaysHere agree that any dispute, claim, or controversy arising out of or relating to Guardian or to these Terms — including their formation, interpretation, breach, or termination, and including claims in contract, tort, statute, or any other theory — will be resolved by final and binding arbitration, not in court, except as stated in Sections 19.4 and 19.5.

• Who arbitrates. The arbitration will be administered by the American Arbitration Association (AAA) under its Consumer Arbitration Rules, by a single arbitrator. Those rules and the AAA’s consumer fee schedule are at adr.org. If the AAA is unwilling or unable to administer the arbitration, you and AlwaysHere will agree on a substitute administrator, or a court may appoint one.
• Where. If you live in the United States, the arbitration will take place in the county where you live, or by telephone, videoconference, or documents only, at your choice. You will never be required to travel to Minnesota to arbitrate.
• Cost. Because Guardian is free and you are a consumer, AlwaysHere will pay all AAA filing, administrative, and arbitrator fees for any claim you bring, other than the initial consumer filing fee set by AAA’s schedule, which we will also reimburse on request unless the arbitrator finds your claim was frivolous. Each party otherwise pays its own legal costs, except that the arbitrator may award costs and attorneys’ fees where a statute allows it — including, where it applies, the Minnesota Private Attorney General statute.
• What the arbitrator can do. The arbitrator has the authority to decide their own jurisdiction, to rule on the scope, enforceability, and validity of this arbitration agreement, and to award any relief a court could award on an individual basis, including statutory damages and individual injunctive relief. The arbitrator’s decision is final and may be entered as a judgment in any court with jurisdiction. The arbitrator must issue a written award stating the essential findings and conclusions.
• What we do not take away. This section does not limit your right to be represented by a lawyer at any stage, does not shorten any statute of limitations, does not waive any ground for vacating or modifying an award, and does not prevent either party from seeking provisional remedies from a court before an arbitrator is appointed. It does not prevent you from reporting anything to a government agency or from participating in an agency’s own proceeding.

19.3 Your right to opt out — 30 days

You can reject this arbitration agreement and keep your right to go to court.

Email legal@alwayshere.app with the subject line ARBITRATION OPT-OUT, from the email address on your Guardian account, within 30 days of the date you first accept these Terms. Include your name and account email. That is all it takes. There is no form and no phone call.

If you opt out, Section 19.2 does not apply to you and Section 19.6 governs your disputes instead. Opting out has no other effect. It does not affect your account, your access to Guardian, or anything else in these Terms, and we will not treat you differently for doing it.

If we later change this arbitration agreement in a way that is material to you, you get a fresh 30-day opt-out window from the date that change takes effect.

19.4 Class-action and jury-trial waiver

Arbitration under this section is individual only. You and AlwaysHere each agree that claims will be brought only in an individual capacity, and not as a plaintiff or class member in any class, collective, consolidated, coordinated, or representative proceeding. The arbitrator may not consolidate more than one person’s claims and may not preside over any form of class or representative proceeding.

You and AlwaysHere each waive the right to a trial by jury for any dispute covered by this section.

If this class-action waiver is found unenforceable as to a particular claim, then that claim — and only that claim — must proceed in court under Section 19.6, and the remaining claims stay in arbitration. If this waiver is found unenforceable in its entirety, then the whole of Section 19.2 is void and all disputes proceed under Section 19.6.

19.5 Carve-outs — what does not go to arbitration

The following are excluded from Section 19.2:

• Small claims. Either party may bring an individual claim in small claims or conciliation court, so long as the claim stays in that court, stays individual, and is within that court’s jurisdictional limits.
• Injunctive and equitable relief. Either party may go to court for a temporary restraining order, preliminary injunction, or other provisional or equitable relief — for example, to stop misuse of install keys, to stop covert monitoring in breach of Section 6, or to protect intellectual property or someone’s safety.
• Claims that cannot lawfully be compelled to arbitration. Including any claim alleging sexual assault or sexual harassment, which you may bring in court at your election under the federal Ending Forced Arbitration of Sexual Assault and Sexual Harassment Act, and any claim for public injunctive relief where the law reserves it to the courts.

19.6 Court, where arbitration does not apply

Where a dispute is not subject to arbitration — because you opted out, because it falls under Section 19.5, or because Section 19.2 is held unenforceable — it will be brought exclusively in the state or federal courts located in Hennepin County, Minnesota, and you and AlwaysHere each consent to the personal jurisdiction of those courts. The small-claims carve-out in Section 19.5 is an exception: you may bring a small claim where you live.

19.7 Survival

This Section 19 survives termination of these Terms and closure of your account.

20. General

• Entire agreement. These Terms and the Guardian privacy policy at guardian.alwayshere.app/guardian-privacy are the whole agreement between you and AlwaysHere about Guardian, and replace any earlier understanding — including anything stated on our marketing pages. Where a marketing page and these Terms conflict, these Terms control.
• Severability. If any provision is held unenforceable, it is modified to the minimum extent needed to make it enforceable, or severed, and the rest stays in force.
• No waiver. If we do not enforce a provision, we have not waived it.
• Assignment. You may not assign these Terms. We may assign them to a successor in connection with a merger, acquisition, or sale of assets.
• No third-party beneficiaries. These Terms create no rights in anyone other than you and AlwaysHere. The Supported Person is not a party to this agreement and is not bound by it, including by Section 19. Nothing here limits any right the Supported Person has under applicable law, including any right to object to being monitored.
• Force majeure. We are not liable for failures caused by events outside our reasonable control, including outages at Anthropic, Twilio, our hosting providers, or the mobile networks.

21. Contact

AlwaysHere LLC
584 Hyacinth Place, Highland Park, IL 60035
legal@alwayshere.app

Use that address for anything legal or formal, including:

• CONSENT in the subject line — to report that Guardian was installed on someone without their consent.
• ARBITRATION OPT-OUT in the subject line — to opt out under Section 19.3.
• DELETE in the subject line — to delete your account and data under Section 14.

These Terms live at guardian.alwayshere.app/guardian-terms. The Guardian privacy policy is at guardian.alwayshere.app/guardian-privacy.

Guardian is a product of AlwaysHere LLC. Claude is a product of Anthropic, PBC, which is not affiliated with AlwaysHere and does not endorse Guardian.