Privacy / VEYLUM
Local first by design
Installed ARC and imported GGUF models run on your computer. Local chat requires no account. If you explicitly choose an OpenAI API connection and approve an answer, its chat context, instructions and selected memories go to OpenAI. Project tools and background tasks cannot use that connection. API billing is separate from ChatGPT subscriptions.
Chats, bots, memory, project context, task state, and settings are stored on your desktop. Attached workspace files remain in the folders you choose.
The application does not send analytics or telemetry. This website has no trackers, external fonts, third-party scripts, or address storage.
Installing a model contacts its configured third-party source. Optional mobile access serves your paired device over the private LAN when you turn it on. Public Hugging Face model downloads contact Hugging Face. Approved API answers contact OpenAI. Binary hosting sees ordinary download requests.
Private data still needs care
Memory and project scope determine what local context a conversation may use. Inspect and remove memories when they are no longer helpful.
Local storage is not a promise of encryption at rest. Your Windows account, device encryption, backups, and anyone with access to your desktop affect its privacy. Plain .veylum backups contain private information and are not encrypted. Optional passphrase protection encrypts an exported archive; a lost passphrase cannot be recovered. The live database and rollback copies remain unencrypted. Protect private files and never upload it as a public release asset.
Veylum Mobile connects to your own desktop’s HTTPS host. Pair deliberately and revoke devices you no longer use. The address checker on this site validates text locally; it sends no request to your desktop and does not save the address.
Cloudflare hosts this public website and may process ordinary web requests. Private chats, models, keys, account vaults and desktop data are not uploaded with the site.
Read the security boundaries →