Privacy Policy

This Privacy Policy explains what information Atlas ("Atlas," "we," "us") — a product of Roi Tech LLC — collects, how it is used, and the control you have over it. Atlas is currently in private, early-access development and is provisioned as a separate, single-owner installation for each authorized user.

01Who Atlas is for

Each Atlas installation has a single authorized owner. Atlas acts only on that owner's behalf, using only the accounts and services the owner explicitly connects. There is no shared, multi-user access to a single owner's data.

02What Atlas accesses

To do its job, Atlas connects — with your explicit authorization — to the services you choose to link. Depending on what you enable, that may include your calendar, email, messages, contacts, reminders, location, files, and connected devices such as your home or car. Atlas reads from and, with your approval, acts on these services, and builds an evolving model of the context it needs to help you (your people, places, preferences, routines, and commitments).

• You choose what Atlas can see. Each connection is opt-in, and you can disconnect it at any time.
• You choose what Atlas can do. Actions that change something in the real world — sending a message, booking a reservation, controlling a device — require your approval.
• Credentials stay server-side and encrypted. Access tokens and account credentials are stored encrypted and are never exposed in the interface.

03Feedback: screenshots & chat excerpts

Submitting a screenshot or chat excerpt with your feedback is always optional. If you'd rather not include potentially sensitive content, you can describe the issue in words instead.

04How your data is used

Your data is used to operate Atlas for you and no one else: to understand your requests, remember relevant context, act on your behalf with your approval, and proactively surface things that need your attention. Atlas does not sell your personal data, and does not use your personal data to train third-party AI models.

Atlas relies on third-party model providers (such as Anthropic, OpenAI, and Google) to process requests. Content sent to those providers is governed by their respective terms; Atlas is configured to use these services in a way that does not contribute your content to their model training where such controls are offered.

05Accountability & traceability

Every action Atlas takes on your behalf can be traced. You can review what Atlas did, revoke any connection, and remove stored context. Because important actions are gated behind your approval, nothing consequential happens without your say-so.

06Data retention & deletion

Your context and connected data persist so Atlas stays useful over time. You can delete stored memories, disconnect any account, or request full deletion of your installation's data. On disconnection, Atlas stops accessing that service and removes the associated credentials.

07Security

Atlas is designed around the principle that an assistant this capable must be safe to trust: single authorized owner, server-side encrypted credentials, human-in-the-loop approval for consequential actions, guarded handling of untrusted content, and an auditable record of activity. No system is perfectly secure, but control and traceability are built into how Atlas works.

08Changes to this policy

As Atlas evolves, this policy may be updated. Material changes will be reflected here with a revised "last updated" date.

09Contact

Questions about privacy, or a request to access or delete your data, can be sent to atlas@roitechllc.com.