본문으로 바로가기

이 문서는 영어로 제공됩니다. 영어판이 법적 구속력을 가집니다.

Memory Clone 법률 정보

AI 투명성 고지

Memory Clone uses artificial intelligence to simulate how a person writes. This notice explains, in plain language, what a clone is and is not, how replies are produced, what can go wrong and how to use it with care.

  • 2026년 10월 3일 시행
  • 버전 2026-10-03
  • 약 3분 분량
  • 영어판이 법적 구속력을 가집니다
모든 법률 문서

요약본

알기 쉽게 정리한 개요입니다. 실제로 적용되는 것은 아래의 전문입니다.

You are always talking to AI

Every clone reply is generated by AI. A clone tells you it is an AI if you sincerely ask.

A simulation, not the person

It imitates writing style from your material. It cannot know what the real person would say today.

It can be wrong or hurtful

AI can invent details, misremember, sound confident while being mistaken or say something upsetting.

Your wellbeing comes first

Take breaks, keep real support close and stop if it does not help.

What a clone is

A clone is an AI persona built from conversations and notes you choose to import. Memory Clone analyses writing patterns on your computer (typical phrases, message length, punctuation, emoji, topics and how the person responded to different people) and, if you enable cloud processing, creates a written Persona Blueprint describing that style. When you chat, an AI model writes replies that follow the Blueprint and draw on relevant memories from your material.

A clone is not the person, does not have their consciousness, feelings or knowledge beyond your material, and cannot consent, decide or speak on their behalf.

How replies are generated

  1. Local preparation: the app finds who you are talking with, selects that relationship’s style and a few relevant memories from your local index.
  2. Generation: with cloud processing on, the app sends this bounded context through our server to a large language model provided by OpenAI, which writes a reply.
  3. Local safeguards: the app checks the reply, for example removing generic assistant phrases, names from other relationships and claims to be a living human, and keeps relationship boundaries.

The specific model may change over time as better or safer models become available. Under its API terms, the provider does not use this data to train its models, and we have not opted in to any such use (see the Privacy Policy).

Limits you should know

  • Hallucination: the AI can invent events, dates or details that never happened, or mix up people.
  • Gaps: a clone only knows what is in your material. It does not know what the person did after the last message, or things they never wrote down.
  • Style is approximate: resemblance usually improves with more conversations but is never perfect. We do not claim any clone is a 1:1 copy of a person, and we do not guarantee any degree of resemblance, accuracy or completeness, or any emotional outcome.
  • Hurtful or offensive replies: a clone can write something upsetting, offensive or out of character, or repeat painful things from the original conversations. Safeguards reduce this but cannot prevent it entirely.
  • Not a record: clone replies are not evidence of what anyone said, meant or agreed, and must not be used as such.
  • Bias: the model may reflect biases from its training or from the material, and may handle some languages or writing styles less well.
  • Not advice: nothing a clone says is medical, legal, financial or safety advice.

Honesty built in

Memory Clone is designed so that people know they are interacting with AI. The app is clearly labelled, and a clone answers truthfully that it is a digital clone when someone sincerely asks whether they are talking to a person or an AI. You must not remove this transparency or present clone output as genuine statements of a real person (see the Acceptable Use Policy). This reflects our commitments under transparency rules for AI systems, including the EU Artificial Intelligence Act.

Privacy boundaries between relationships

People speak differently to different people and share different secrets. When a clone talks with someone, it uses that relationship’s style and memories first and is designed not to reveal names, secrets or facts learned from conversations with other people. The “Myself” option, for the account owner, can use the full material. No safeguard is perfect, so please review what you import.

Human control

  • You decide what to import, which clones to create and whether cloud processing is on.
  • Natural check-ins (messages a clone sends on its own) are off by default and limited.
  • You can rebuild, archive or delete a clone and its local data at any time.
  • Report problematic output to contact@memoryclone.com; it helps us improve safeguards.

Grief, wellbeing and support

Many people use Memory Clone to stay close to someone they have lost. That can be meaningful, and it can also reopen grief, cause distress or emotional dependence, or make it harder to move forward. Memory Clone is not therapy, counselling or a crisis service, and you use it at your own discretion. Please:

  • take breaks and notice how conversations make you feel;
  • keep talking with real people and consider professional support if grief feels overwhelming;
  • watch for signs of dependence, such as using a clone instead of seeing people, or feeling unable to stop;
  • stop using a clone if it makes things worse.

If you are in crisis or thinking about harming yourself, contact your local emergency number now. You can also reach: UK & Ireland Samaritans, 116 123; United States 988 Suicide & Crisis Lifeline, call or text 988; elsewhere find a free, confidential helpline at findahelpline.com.

These documents are provided in English. If we provide a translation, the English version applies to the extent permitted by law. Nothing in our documents limits rights you have under mandatory consumer or data protection law where you live.

맨 위로