본문으로 바로가기

이 가이드는 현재 영어로만 제공됩니다.

프라이버시, 동의 및 신뢰

A Private, Local-First AI Chatbot for Personal Messages

Personal messages are among the most sensitive data anyone has. Before you hand them to an AI, it is worth knowing exactly where they go. This guide explains local-first design in plain language, including its honest limits.

  • 업데이트
  • 약 4분 분량
  • By the Memory Clone team

간단한 답변

A local-first AI chatbot keeps your original files and searchable memory index on your own computer and sends only what is needed to generate a reply. In Memory Clone, imports and the index stay on your Windows PC; cloud AI is used only after you enable cloud processing for a clone, and your data is not used for training.

Cloud-first vs local-first, in plain terms

In a cloud-first AI service, you upload your files to the company’s servers, where they are stored, processed and searched. Convenient, but your entire archive now lives somewhere else.

In a local-first design, your files and the work done on them stay on your device by default. The app only reaches out to the internet for things that genuinely need it, such as your account and licence, and, if you choose, generating a reply with a large AI model.

QuestionTypical cloud-firstMemory Clone (local-first)
Where are original chat files kept?Provider’s serversYour Windows PC
Where is the searchable memory index?Provider’s serversYour Windows PC
What is sent to generate a reply?Often already stored in the cloudOnly excerpts needed for that reply, once you enable cloud processing
Used to train AI models?Depends on the providerNo, Memory Clone does not train models on your data
What the server storesYour contentAccount, licence, plans and credit usage; chat message content is not stored

What happens to your data in Memory Clone

  1. ImportYou export chats yourself and import the files. Parsing happens on your PC; originals are not uploaded.
  2. IndexThe app builds a local memory index and style profile for each relationship, stored in your workspace on the PC.
  3. Consent to cloud processingBuilding the clone’s Persona Blueprint and chatting need a large AI model. Nothing is sent until you switch on cloud processing for that clone.
  4. Generate a replyFor each message, the app selects the relevant memories and style examples and sends those excerpts through the Memory Clone server to the AI provider (OpenAI), which returns a reply.
  5. Bill usageThe server records credit usage for your account. It is the source of truth for credits so that they cannot be spent twice.

Relationship boundaries are also a privacy feature. When the clone talks to one contact, it is designed not to reveal names, secrets or facts that only appeared in conversations with someone else.

Honest limits of local-first

  • Replies use the cloud. Running a model as capable as current cloud AI on an ordinary PC is not practical yet, so chatting requires sending excerpts once you enable it.
  • Your device becomes the vault. Local storage is only as safe as the computer. If someone can use your Windows account, they can reach your workspace.
  • Backups are your job. Data that lives only on your PC can be lost with it, so keep your original exports backed up securely.

How to protect a personal message archive

  1. Use a Windows account with a strong password or PIN, and lock the screen when you step away.
  2. Turn on disk encryption (BitLocker or Windows Device Encryption, where available) so a lost or stolen laptop does not expose your files.
  3. Keep Windows and the app updated.
  4. Store original exports in one known folder rather than leaving copies in email, downloads and shared drives.
  5. Do not share your PC account with people who should not see the clone or its sources.

Questions to ask any personal AI app

  • Are my original files uploaded, or processed on my device?
  • Is anything sent before I explicitly agree?
  • Is my data used to train models, by the app or by its AI providers?
  • Can I delete my data, and what remains on servers if I do?
  • Does the app keep different people’s secrets separate from each other?

Our Privacy Policy answers these for Memory Clone in detail, and the AI Transparency Notice explains how replies are generated.

Respecting the people in your chats

Privacy is not only about you. Your chats contain friends’ and family members’ words. Keep clones private, do not use them to expose what others said, and get permission before cloning another living person. The Acceptable Use Policy sets out the rules.

Frequently asked questions

Can I use Memory Clone completely offline?

Importing and indexing happen on your PC, but chatting uses cloud AI and an internet connection for your account and credits.

Does Memory Clone upload my WhatsApp chats?

No. Original files stay on your PC. With cloud processing on, only the excerpts needed for each reply are sent to generate it.

Is my data used to train ChatGPT or other models?

Memory Clone does not use your data to train models. Excerpts are sent to the AI provider only to generate replies.

What does the Memory Clone server store?

Your account, licence, plans and credit usage. It passes excerpts to the AI provider without storing chat message content; a delivered Persona Blueprint is kept briefly so retries are not charged twice. The Privacy Policy has the details.

What happens if I switch cloud processing off?

The clone stops generating new replies until you turn it back on. Your local data stays on your PC.

App menus change over time, so labels on your device may differ slightly. Memory Clone creates an AI simulation from the material you provide; it is not the person.

내 메시지로 직접 체험해 보세요

Memory Clone은 Windows 앱입니다. 가져온 채팅과 기억 인덱스는 내 PC에 남아 있으며, 클라우드 답변은 클론의 클라우드 처리를 켠 후에만 시작됩니다. 만 18세 이상이어야 합니다.

맨 위로