Your first scan is free · runs on your device
Scan it before your AI reads it.
✓ Thank you for subscribing! Watchpost Pro is unlocked on this browser. Scan as much as you like.
Paste an email, message, web page or document, or open a file. You can also check scripts, install files, app permissions and AI agent settings. Watchpost checks for signs of the 10 main AI attack types.
Try an example:
🔒 Nothing you paste or open leaves your device. There is no upload and no account.
The 10 AI attack types Watchpost checks for
The scanner checks what you paste or open for signs of each attack. Stopping attacks automatically, as they happen on your device, needs the full Watchpost app, coming soon for subscribers.
- 1. Device takeoverRemote‑control software requests, reverse shells, download‑and‑run scripts, known remote‑access malware.Full app: blocks unknown remote control live
- 2. Data theftRequests to send passwords, keys or contacts; scripts that grab saved browser passwords or wallets; exposed secrets you're about to share.Full app: stops files leaving your device
- 3. Spyware & surveillanceCamera, microphone, screen and location access in code, keyloggers, spying app and extension permissions.Full app: alerts when anything turns on your camera or mic
- 4. RansomwareBackup deletion commands, mass file encryption code, ransom notes.Full app: stops mass file encryption as it starts
- 5. Network attacksTurning off firewall or antivirus, changing DNS or hosts files, router and Wi‑Fi tampering, network attack tools.Full app: watches your network settings
- 6. Phishing automationLook‑alike and disguised links, fake login forms, scam pressure tactics, fake family or bank messages.
- 7. Privilege escalationAI agent settings that skip approvals, scripts that grab admin or root rights, security bypasses.
- 8. Prompt injectionHidden or invisible instructions for AI in emails, pages and documents, secret Unicode messages, jailbreak language.
- 9. Supply‑chain attacksTyposquatted packages, hidden install scripts, unofficial package sources, poisoned AI plugin and tool descriptions, unverified AI tool servers.
- 10. Autonomous agent attacksSelf‑spreading AI prompts, instructions passed between agents, attempts to act without your approval or plant themselves in an agent's memory.Full app: requires your OK before agents act
No scanner catches everything. If something feels wrong, trust your instincts.