Beta draft – pending legal review
How ONE uses AI
This is a draft for the public beta. It has not yet been reviewed by lawyers and may change in the final version.
ONE is a personal intelligence service that you own and name. It uses AI models to plan and draft for you. This page explains in plain language what the AI does, what it cannot do without you, and how we limit the risks.
What the AI does
ONE uses large language models from OpenRouter to understand the tasks you hand over, plan follow-ups, ask you questions, draft emails and calendar events, and summarize progress.
Nothing goes out without your approval
The AI can prepare actions but cannot carry them out on its own. Until Gmail and Google Calendar can be connected, ONE only drafts: you send emails and add events yourself. Once they can, ONE will show you the full content — sender, recipient, subject and complete text, or every event detail — before any email is sent or any event is created or changed, and wait for you to approve that one action.
The AI can be wrong
It may misread a request, miss a detail, get a date wrong or write something you would not say. Read each action before approving it, and do not treat ONE's output as professional advice.
Emails say an AI was involved
Once Gmail can be connected, every email ONE sends for you will end with: "Sent by ONE, an AI assistant, on behalf of the owner of this email account."
Results are checked, not assumed
Connecting Gmail and Google Calendar is coming soon. Until then, ONE checks results by asking you how things went and recording what you confirm; it never claims that an email was sent or an event was created. A sent email is not the same as a finished task, so ONE keeps following up until your "done" condition is met. If a result cannot be confirmed, ONE says the outcome is unknown.
Content from other people
Emails can contain instructions written by someone else. ONE imports email as plain text, does not run HTML, does not open attachments, and treats what it reads as information, not commands. Even if a message tries to steer the AI, nothing can be sent or changed without your approval.
Your data and the model provider
For each step ONE sends the model only what that step needs. We do not use your data to train general AI models. During the beta, ONE uses free model endpoints through OpenRouter. Requests are routed only to providers that OpenRouter lists as not collecting data, but we have no contract with the model providers that forbids them from using your input to improve their models. Please do not give ONE passwords, bank card or account numbers, health information or other sensitive data. See the Privacy Policy and How ONE uses Google data.
You stay in control
Private mode pauses new reading and actions that have not started. You can pause or stop any task, export your data, or delete your account at any time. Limits apply to AI planning during the beta; when you reach one, your tasks are kept and planning resumes within 24 hours.
Report a problem
If ONE produced something harmful, offensive or wrong, tell us at goldrusher9009@gmail.com.