SAFETY BY DESIGN
You stay in control—always, visibly, instantly.
AI Desktop Agent only acts with explicit permission. When it can see your screen or guide clicks, you’ll see clear on-screen indicators, and you can stop it immediately via Esc, Stop, or a mouse move. It confirms risky actions, and it never signs in or types passwords—so automation stays safe, transparent, and user-led.

Safety by design, not by promise
AI Desktop Agent only acts when you ask. Every mode uses clear on-screen indicators, explicit permission prompts, and instant stop controls, so you can guide clicks, review actions, or automate tasks while staying in charge at every step.
- Always asks first before viewing your screen or acting in apps
- Visible border/banner plus clear control state, with instant stop via Esc/Stop/mouse
- Confirms risky actions and never signs in or types passwords
SAFETY PRINCIPLES
Three pillars that keep you in control
AI Desktop Agent is built for deliberate, permission-based help. These safety pillars guide every mode—so you always know what it can see, what it can do, and how to stop it instantly.
Permission first, always visible
It only views your screen when you ask. When it’s active, clear borders/banners and control indicators show you exactly what’s happening—no hidden automation, no surprises.
You can stop it instantly
One action ends control right away: press Esc, hit Stop, or move your mouse. The agent yields immediately so you can take over at any moment.
No passwords, no silent sign-ins
The agent won’t sign into accounts or type passwords. For risky steps, it asks for confirmation first, making sensitive actions explicit and easy to decline.
SAFETY FIRST, ALWAYS
Automation you can see, control, and stop
AI Desktop Agent never acts silently. It asks before viewing your screen or doing anything in an app, then shows clear on-screen indicators so you always know when it’s observing or taking control. Use Idle mode for chat only, Co-Pilot for step-by-step click guidance, or Pilot for assisted action—only with your permission.
You stay in charge: stop instantly with Esc, the Stop button, or your mouse. The agent confirms risky steps and won’t sign in or type passwords. Prefer privacy-first setups? Run local models with Ollama and use on-device speech recognition options when available.


SAFETY METRICS
Controls built for confident, interruptible automation
100%
ACTIONS REQUIRE EXPLICIT PERMISSION
3
INSTANT STOP OPTIONS: ESC, STOP BUTTON, OR MOUSE
0
PASSWORDS TYPED OR AUTO-SUBMITTED BY THE AGENT
5
VISIBLE MODE STATES: IDLE, CO-PILOT, PILOT, AUDITOR, TUTOR

SAFETY BY DESIGN
You stay in control—always, instantly.
AI Desktop Agent only sees your screen or takes action when you explicitly ask. Clear on-screen indicators show when it’s viewing or driving, and you can stop immediately with Esc, the Stop button, or your mouse—no waiting, no debate.
For risky steps, it asks for confirmation and won’t sign in or type passwords. You can keep things private with local model options and on-device speech recognition, while still getting help grounded in official manuals with cited sources.
NEXT STEP
Want to validate our safety controls in your environment?
Talk to us about your workflows and risk requirements. We’ll walk through permissions, visible indicators, and instant stop behavior—no hard sell.