Security overview

Desktop-first trust boundary. Hosted where it counts.

App, backend, update feed, and provider routing all matter.

Threat model highlights

  • Local machine compromise can expose local caches, screenshots, or sessions.
  • Hosted service compromise can affect accounts, billing, or update delivery.
  • AI output can still be wrong or unsafe even when infrastructure is healthy.
  • Browser automation and computer-use features should be treated as powerful actions that deserve clear user intent.

What Sylica AI does today

  • Uses HTTPS-backed hosted infrastructure for web access and update delivery.
  • Separates the desktop app from the backend and publishes versioned Windows update artifacts.
  • Keeps the app mostly desktop-driven so the full interaction loop does not depend on the public website UI.
  • Allows browser automation to pause for sensitive inputs rather than blindly auto-filling every step.

User guidance

  • Install only builds published from the official domain.
  • Keep Windows, the browser, and the desktop app updated.
  • Review automation tasks before use, especially browser actions with billing or account impact.
  • Do not rely on AI output alone for legal, security, or compliance decisions.