Security

Handing an AI agent on a server somewhere you don't know seems like a big security and privacy concern. This document serves to show full transparency so you know where your data is.

Server information

This agent runs on an Ubuntu box located in Miami, Florida.
I rent it through Linode/Akamai technologies.

API information

Talon runs on the OpenAI API (GPT 6 Luna, and GPT 6.1 Sol on Enterprise) and the TypeSafe API (Jev).
These companies have their own privacy policies, but they are sent any message you send and can see data returned from tools.

Your data

Your credentials for websites are saved as cookies on this server. They are never sent to OpenAI or TypeSafe, and are encrypted on this server.

Browser proxies

Your agent's browser reaches websites through proxy servers from Webshare, so sites see a Webshare address rather than this server's. Secure (https) pages stay encrypted end to end: Webshare can see which site is visited and when, not what's on the page, your logins or what's typed. The proxy passwords stay on this server.
Composio and other MCP servers have their own privacy policies regarding sign-ins for connecting apps.
The site sits on top of Cloudflare and all traffic goes through them.
Any messaging service you use, such as Discord, will be able to see your message history within the app.
Your files and conversations are saved on this server until you delete them.

Deletion

Anything you delete, whether it is files or conversations, will be deleted from this server automatically and immediately.
However, this server is backed up, so your files may sit in an archived backup for up to 14 days.
OpenAI and TypeSafe have their own policies regarding this.

About me

My name is Elliot DeBruin and I am the sole proprietor/owner of this business. I am the only one with credentials to the server and the only person who has ever logged in to it.