AI companies spent years getting users to paste data into chat boxes. Now they do not ask. They take pictures of your screen every few seconds and read those instead.
This is not theoretical. Anthropic's own documentation for Claude Desktop's "computer use" feature states openly that it works by "capturing screenshots and sending them to the model for analysis." Microsoft's Copilot Vision does the same. Dozens of agent tools that claim to "use your computer like a human" operate on the same principle.
The shift from text to vision happened fast. The privacy implications are severe.
Text Was Controllable
When ChatGPT was a browser tab, you decided what to paste. A code snippet, an email, a document. The AI never saw your Slack, your password manager, your bank balance, or that DM you regretted. The boundary was clear: you give, they receive.
Local tools that read files followed the same contract. You pointed them at a directory. They read those files. They did not open your email client and read your inbox.
That contract is gone.
Screenshots Are Everything
A screenshot does not discriminate. It captures your entire display in a single frame. When an AI agent screenshots to "see" what you are doing, it sees:
- Your password manager autofilling credentials
- Your Slack DMs in the sidebar
- Your bank balance in the background tab
- The confidential document on monitor two
- The email from your lawyer
- The calendar event titled "Interview at Competitor — 2pm"
- Your terminal with API keys in plain text
- The spreadsheet with employee salaries
All of it. In one image. And most users have no idea because it is framed as "the AI can see your screen to help you."
The "Local" Lie
"Local AI" sounds private. The model runs on your machine. Your data does not go to OpenAI's servers. Great.
But the screenshot still happens. The vision model still processes the image. And if the agent has internet access — which most do, because otherwise it is useless — that processed information can be logged, synced, or sent to a cloud dashboard for "improving the agent." Local inference does not mean local observations.
Some tools upload screenshots to their servers for "session replay" so you can review what the agent did. A third-party server has a photo album of everything you did while the agent ran. Every window. Every tab. Every notification.
Worse Than Screen Sharing
In a Zoom call, you know you are sharing. Green border. Stop Share button. Preview pane. You chose to broadcast, and you can end it.
AI agents do not announce when they screenshot. No border. No indicator. No preview. They capture in the background while you work, often every two to five seconds, and you have zero visibility into what they observed or extracted.
With screen sharing, you are holding the remote. With AI agents, you are not even in the room.
Blocking It
These agents use the exact same capture APIs as Zoom, Teams, and OBS. DXGI Desktop Duplication. BitBlt. DWM compositor access. The same paths NoCapture blocks at the OS level.
When you cloak a window with NoCapture, it does not just hide from Zoom. It hides from everything that captures the screen — including AI agents. The compositor omits the cloaked window from the frame buffer. The vision model cannot see it because the pixels are never generated in the capture output.
This is not a feature we added for AI. It is the same protection we have always provided. But the threat model changed. The attacker is not just a meeting you forgot to leave. It is an AI agent running 24/7, taking photos of your life, feeding them to a model that never forgets.
We wrote about the full mechanics of screen capture APIs here.
What to Actually Do
If you run local AI agents — and you probably will — think about your screen like your camera. You would not aim your webcam at your password manager. Do not aim your screen at an AI that screenshots constantly.
- Do not run screenshot-hungry agents while sensitive apps are open. This works but breaks your workflow.
- Use separate machines or VMs? Overkill. The agent is supposed to help with your actual work, not a sanitized sandbox.
- Cloak sensitive windows at the OS level. Keep your password manager, Slack, banking app, and notes locally visible and usable. The AI agent's screenshots simply do not include them. The vision model sees your IDE. It does not see your 1Password. It sees your browser. It does not see your terminal with production credentials.
We wrote about how to hide apps without breaking workflow here.
The Bottom Line
AI companies moved from "ask for text" to "photograph everything" because it is easier. APIs are hard. Integration is hard. Screenshots are trivial. But that convenience costs your privacy — and most users do not realize they have granted camera access to their entire digital workspace.
Your screen is where you actually live and work. It is not a public kiosk they are entitled to photograph. If an AI is going to act like it is streaming your life, start treating it like one.
NoCapture blocks AI agent screenshots the same way it blocks Zoom, Teams, and OBS — at the OS compositor level, before pixels reach the capture tool. Free for two windows. Because "local AI" should not mean "local surveillance."
NoCapture provides OS-level window cloaking, title masking, and notification shielding for Windows. Free for two windows. Because "being careful" isn't a strategy.
