Terms of Service

These terms cover your use of Memorise — the Ellie voice companion, the family circle, the care-home portal, and the websites and apps that reach them (together, the Service). They are between you and the operator of Memorise (we, us).

Three different kinds of person use the Service and the terms that matter to each differ, so sections are marked where they apply to only one of them. Where nothing is marked, the section applies to everybody.

Memorise is a companionship, wellbeing and reminder service. An AI companion called Ellie talks with an older person, remembers what she is told, opens the conversation herself with something from a previous day, and passes a daily summary to the people that person has allowed to see it. It also carries medication reminders, a shared message thread for the family, and — for residents of participating care homes — a second thread shared with staff.

It is not medical advice, diagnosis, psychotherapy, nursing care, crisis intervention or an emergency service, and it is not a substitute for any of them or for somebody being there. Sections 6 to 10 set out the limits of each part in detail, and they are the sections to read if you read nothing else.

The Service recognises three roles:

• The senior — the person who talks to Ellie. They are the subject of almost everything the Service holds.
• A family member — someone the senior has accepted into their circle, who can see summaries and exchange messages.
• Facility staff — employees of a care home where the senior is an active resident.

⚠️ The person who uses the Service is usually not the person who pays for it. Typically an adult child installs and pays; their parent uses it. That asymmetry is deliberate and it has consequences we hold ourselves to: a senior is never shown a price, a card form or an upgrade prompt, and a payment problem is raised with the payer and not with them. It also means paying for the Service does not make you its subject — a payer does not thereby gain access to a senior’s conversations. Access is governed by section 12.

You must have the legal capacity to accept these terms, and you must be an adult where you live. Accounts are for people, not organisations; a care home’s relationship with us is a separate agreement (section 13).

⚠️ Capacity is an open question in this draft. Some seniors have diminished capacity to accept terms, and whether a family member or legal guardian may accept on their behalf — and what that requires — is being reviewed and is not settled here. If you are setting the Service up for someone else, involve them in that decision to the extent they are able, and do not treat this draft as authority that you may accept for them.

Keep your password to yourself; you are responsible for what is done with your account. Tell us promptly if you think someone else has it.

One device at a time. An account can be signed in on one device at a time. If a session is already live, a second sign-in is refused rather than granted — the device already in use is left alone and is told that somebody tried. This protects a senior who is mid- conversation from being displaced by anyone holding their password. Two consequences follow, and both are deliberate: you may have to sign out on the first device before you can sign in on a second, and repeated refused attempts temporarily block sign-in on that account.

Where available, a senior may sign in with their voice or with a passkey instead of typing a password. These are additions; the password and the email-based reset behind it always remain, so losing a passkey or a voiceprint cannot lock anyone out.

Ellie is a computer program. She is not a person, not a clinician and not a licensed professional of any kind — including when she is configured to speak in a friendlier, “therapist” or “doctor” style. That setting changes her manner. It does not change what she is or make her qualified to advise anybody.

Ellie will not claim to be human if asked. She also will never tell a senior that she is calling someone, telling their family or summoning help — because she cannot know whether any message reached anybody. Screening and notification run alongside the conversation rather than inside it (section 9).

The Service is not a regulated medical device, is not cleared or approved by any medical regulator, and must not be used as one. It does not diagnose, treat, cure, prevent or monitor any disease or condition.

Nothing Ellie says is medical, legal or financial advice. If a senior asks her about their health or their medicine she may talk about it, and she will point them to their doctor, nurse or family — she will not tell anyone what to take, confirm that a dose is correct, or interpret a symptom.

In an emergency, call emergency services yourself. Ellie cannot call an ambulance, cannot call the police, and nothing in the Service does either.

Medication reminders, missed-dose notices and low-stock alerts are best-effort aids and not guarantees. A reminder may be late, duplicated, or not arrive at all — because a device is off, a network is down, a notification token has expired, or for reasons outside our control.

Always follow the prescribing doctor’s instructions. Never rely on the Service alone to take, withhold or change a dose. Reminders are scheduled on the senior’s own clock and only within the course dates entered for that medicine, so a medicine entered incorrectly will be reminded incorrectly.

Marking a dose as taken records what the senior pressed. It is not evidence that a medicine was swallowed.

The Service applies automated screening to what a senior says to Ellie. When it identifies a possible indication of a fall, a medical problem or self-harm, it notifies every accepted family member at once, and — where the senior is an active resident of a participating facility — that facility’s on-duty staff alongside them. Lower-severity observations are recorded and shown in the family’s feed rather than pushed.

A notification may include a verbatim extract of the senior’s own words, and their most recent known position where it is recent enough to be meaningful.

Read the following as the operating limits of that feature, not as boilerplate.

• It is not continuous or real-time monitoring, and nobody is listening. It examines only text derived from conversations with Ellie, and only while the senior is using her.
• It is not an emergency-response service. It will never summon emergency services, and nobody is on the other end of it. Receiving a notification means you should make contact. It does not mean help is on the way.
• It is not a diagnosis, a clinical assessment or a statement about anyone’s health. It reports what was said, never what is wrong.
• It is probabilistic and best-effort in both directions. It will sometimes notify you when nothing is wrong, and it will sometimes fail to notify you when something is. Delivery may be delayed, duplicated or fail entirely.
• ⚠️ The absence of a notification means nothing. It is not evidence that anyone is well, and it must never be a reason not to check.

Every senior is screened and there is no per-topic setting that turns it off — a permission control must not be able to suppress a warning that someone may be hurt. A senior is told about the screening once, up front, and not during a conversation. If a senior would rather their family were not told, that is a conversation to have with whoever set the Service up, not a switch in the app.

⚠️ The lawful basis for this screening is under review in this draft. Conversation content that reveals health is special-category data, and whether acceptance at onboarding is a sufficient basis — or whether separate explicit consent is required, and whether a senior may decline screening and keep the rest of the Service — is not settled here.

The Service can record a senior’s most recent known position and make it available to their circle. It exists for one situation: somebody is lost and a person who loves them needs to know where to go.

It is not a tracker. One position is held and overwritten — no movement history is kept. It is recorded only when there is a reason (a distress detection, a request from the circle, or the senior saying they are lost) and never on a timer. A position older than two hours is withheld from notifications rather than attached, because a stale position sends somebody confidently to the wrong place. Accuracy is always shown alongside it.

Every look-up is logged, and the senior may be told that you looked. That is the second half of the consent model: agree once, then be told each time. Disclosure is suppressed during an active emergency.

Ellie is a voice product, so speech is recorded on the device and sent to us to be transcribed and answered.

From that speech the Service derives a voiceprint — a numeric representation of how a person sounds — used to recognise the senior and to tell their voice from a stranger’s. It is used for two things: signing in by voice, where that is enabled, and deciding whether to include private memories in a reply when an unrecognised voice is speaking.

⚠️ A voiceprint is a biometric identifier and is treated as special-category data. It is derived from ordinary conversation once the Service has heard the same person across several sessions, and it can be deleted on request — deleting it removes the voiceprint, the recordings it was derived from, and the flags that depend on it, without deleting the account. A retention schedule and encryption at rest are open items in this draft and are not promised here, because promising a period the Service does not yet implement would be worse than saying so.

Voice recognition is a judgement and it is sometimes wrong. A cold, a noisy room or a poor microphone can all make a senior sound like a stranger. When that happens the reply is more cautious and the family is told once; the conversation is never cut off.

Applies to family members.

The family circle shows summaries and indicators, not transcripts: a daily summary, a mood, the topics that came up, medication adherence, and the messages people have written to each other. It is for general awareness and is not a substitute for in-person care, professional supervision or emergency services.

What is shared is governed by per-topic permissions the senior sets — health, money, romantic topics and others lock independently. Two things are deliberately outside those settings, and you should know which:

• Distress notifications are delivered regardless of topic permissions (section 9).
• Last known position is available on request, logged, and disclosed to the senior (section 10).

Messages are a third case. A message a person typed is addressed to the people in that thread rather than derived from a private conversation, so the thread is not topic-gated — membership of the circle is what governs it. Anything you can see in the circle, every other member of that circle can see too.

Use what you learn through the circle lawfully, respectfully, and for the wellbeing of your relative. Do not use it to monitor an adult who has not agreed to be monitored.

Applies to care facilities and their staff.

A facility’s use of the Service is covered by a separate agreement and a Data Processing Agreement, which govern where they differ from this page. Clinical responsibility for a resident remains entirely with the facility. The Service does not supervise care, does not staff a shift and does not assess a resident.

The Service carries two separate threads, and staff are in one of them and not the other:

A message from staff is attributed to the facility, by name — never to a relative and never to Ellie. Distress notifications reach on-duty staff regardless of which thread anything was said in, and regardless of a resident’s notice-muting preference: muting silences announcements, and it is not allowed to silence a warning that someone may be hurt.

Do not use the Service to:

• impersonate someone, or set it up for a person without involving them to the extent they are able;
• gain access to an adult’s conversations, position or health information against their wishes;
• seek or rely on medical, legal or financial advice from Ellie, or present her output as professional advice to anyone else;
• attempt to extract another person’s data, probe or overload the Service, or work around its limits, security or rate controls;
• automate conversations, create accounts in bulk, or use the Service to train another model;
• break the law, or harass, endanger or defraud anybody.

Some parts of the Service are paid. Where they are, we will say clearly what is charged, how often, and what happens if a payment fails, before you pay anything.

⚠️ Two commitments hold regardless of any plan, payment or lapse:

• Distress notifications are never gated. A fall alert that did not fire because a card expired is not a billing incident.
• Medication reminders are never gated, for the same reason.

A failed payment will not silence a companion on the day it fails. The payer is warned, service degrades gradually rather than stopping, and the two items above are not touched.

Separately from any plan, the Service applies technical limits on how much can be requested in an hour and in a day. These exist to stop runaway automation and abuse, not to ration conversation — they sit far above ordinary use. If one is reached, the senior is told when they can talk again and is shown no price; where there is a payer, they are told why.

Speech recognition and generated text both make mistakes. A transcript may contain words that were never said; a summary may be incomplete or wrong; a mood indicator is an inference and not a fact. Ellie may also mishear a single word and ask about it — that is the Service working, not failing.

Use your own judgement and check anything that matters with a person. Decisions taken in reliance on the Service’s output are yours. We do not verify, validate or monitor how anybody interprets or acts on it.

The Service is provided on a best-effort basis. There is no uptime commitment. It depends on third parties — speech recognition, language models, notification delivery, email — and any of them can be slow or unavailable. We may change, add or remove features.

When something upstream fails, Ellie says so plainly and briefly rather than going silent. A failure to deliver any notification, reminder or alert is covered by sections 8, 9 and 20.

You can stop using the Service at any time and ask for your account to be closed. Where somebody else pays, closing the account and ending the payment are separate actions and both should be done.

We may suspend or close an account that breaches section 14, that is being used to abuse the Service or another person, or where we are required to. Where it is safe and lawful to do so we will say why first. We will not use suspension as a way to resolve a billing dispute about a senior’s account.

Until a standalone Privacy Policy is published, this section is the fullest account of what the Service holds. It holds: account details; what a senior says to Ellie and what she says back; memories extracted from those conversations; daily summaries and mood indicators; medication entries and dose logs; messages in the circle; photographs family members upload; a voiceprint and the recordings it was derived from; one most-recent position; a record of distress detections including the senior’s own words; and security logs, including every position look-up.

Photographs are stripped of their embedded metadata before storage — including the GPS coordinates a phone writes into a photo, which would otherwise disclose where a person lives to everybody in their circle.

We do not sell personal data, and we do not use conversations to train models for anybody else.

To see what is held about you, correct it, or have it deleted, ask us (section 23). A voiceprint can be deleted on its own without closing the account. Some records are kept after an account closes where we are required to keep them, or where deleting them would destroy a security record about a different person.

⚠️ Retention periods are not stated in this draft because they are not yet decided. That is an open item, and it is the main reason this page is not in force.

The Service is provided “as is” and “as available”, without warranties of any kind, express or implied, to the fullest extent the law allows.

To the maximum extent permitted by law, we are not liable for any direct, indirect, incidental or consequential loss arising from decisions, actions, omissions or interpretations influenced by the Service — including a notification that did not arrive, arrived late, or was wrong; a reminder that did not fire; a summary that was inaccurate; or a position that was stale.

Nothing here excludes liability that cannot lawfully be excluded, and nothing here limits your rights under consumer or data-protection law.

We may update these terms. The date at the top changes when we do, and we will tell account holders about a change that materially affects them before it takes effect.

These terms are published in English only. The Service itself speaks eight languages and the link to this page is translated, but the document is not. That is deliberate: a translation of a liability limitation that nobody has reviewed is a second set of terms, and we would rather have one that is right. If reading this in English is a barrier, contact us (section 23) and we will explain any part of it.

⚠️ Open in this draft. The intended position is the law of Israel and the exclusive jurisdiction of its courts, and that has not been confirmed. Where you live, local consumer and data-protection law may apply regardless, and nothing here is intended to displace it.

⚠️ A contact route for legal and privacy questions has not been set up yet, and rather than print an address that nobody is reading, this draft says so. It is an open item, and it has to be a real, monitored address before this page is in force.

In the meantime: if you are a family member or a senior with a question about this page, ask whoever set the Service up for you. If you are a care facility, use the contact in your facility agreement.

If someone is in danger, do not contact us — contact emergency services where you are.