Skip to content
Homie

Trust & Security

Last updated: September 15, 2026

Useful AI starts with clear boundaries. Homie helps you handle supported tasks. KnoMe helps you control access to the private information those tasks may need.

Current friends pilot

The friends pilot supports the web apps and connected AI clients. Voice, WhatsApp, and background automation are not included in this pilot. Start with a test inbox and review important results; this is an early-access service, not a guarantee of complete detection.

The current Gmail connection requests full Gmail access (https://mail.google.com/) and your email address for account identification. This Google permission is broader than the app’s read-only Gmail tools: the current allowlist permits message, attachment, and profile reads, not sending or deleting email. Permissions in KnoMe further restrict which results an AI application may receive. No contacts permission is requested by this configuration.

Your connection. Your permissions.

Connecting an account should not give every AI application or household member access to it. Our design separates four decisions:

  • Connect: establish a connection to an account you control.
  • Use: allow a supported feature to process the information it needs.
  • Share: allow an AI application or person to receive information.
  • Act: authorize a supported change, message, or other task.

Reading a bill does not authorize paying it. Household membership does not grant access to another adult’s private inbox. The permissions and approval flows available depend on the feature you use.

Privacy by design

We design around limited access, protective defaults, and checks on the server. An AI application’s request does not itself grant authority. Supported access is checked against the relevant account, permissions, and operation.

Privacy filtering can withhold or redact supported sensitive information before a result reaches an AI application. Retrieved messages, documents, and web pages are treated as untrusted content: their contents do not authorize new permissions or actions.

These safeguards reduce risk. They cannot guarantee that every sensitive detail or malicious instruction will be detected.

Who can process your information

For supported Google connections, information moves from Google through Composio to our backend, where the applicable privacy checks are performed before a permitted result is returned to your AI application.

This means our backend and connector provider may process original content. Turning off an AI application’s access does not mean the underlying service processes no information.

Homie also uses AI providers for supported assistance, audio, and document analysis. Saving an extracted record privately does not mean the original was processed without AI. Our Privacy Notice explains the providers, purposes, and information involved.

Private storage and shared spaces

Private information is access-controlled within the service. Information deliberately placed in a shared record or conversation is available to its authorized audience.

We use encrypted connections and storage protections, together with authentication and access controls. Homie and KnoMe are not end-to-end encrypted against our servers and all processing providers. Necessary server-side processing can access content.

Access you can change

Use the available controls to review permissions, revoke a connection, or change what you share. Contact meet@m1factory.com for help or to request access, correction, export, or deletion.

Revoking a permission stops future access through that permission. It does not reverse a completed action or retrieve a copy already received by another person or AI provider. Disconnecting and deleting retained information are separate steps.

AI and your content

We do not use private customer content to train or fine-tune general-purpose AI models. Providers can still retain information for permitted service and security purposes. An AI application you independently choose has its own terms and settings for the copy it receives.

AI responses and scam or authorship assessments can be wrong. Review important details before relying on a result. We do not guarantee prevention of every scam, recovery of losses, or emergency assistance.

Reporting a concern

Email meet@m1factory.com with the affected feature, approximate time, and a short description. Do not include passwords, sign-in codes, or another person’s private information.

We investigate reports, address confirmed issues, and notify affected people and authorities when required. Features and availability may change as the service develops.

Read our Terms of Service and Privacy Notice.