Illuminate. Identify. Control.
Complete AI Endpoint Control
At Lamplighter AI our core mission is to solve all security and privacy risks with using AI endpoint solutions in a business. Our analysis occurs in the browser and on the desktop. It inventories all corporate and shadow AI usage. It uses custom policies to control data movement into AI and all analysis is done on the endpoint, so none of your data is in the cloud.
LamplighterThree questions every security team is now asked about AI.
Which AI tools are people using?
Websites, desktop apps, browser extensions and command-line tools, with who used them, on which device, and for the major apps whether it was a work or a personal account.
What is going into them?
Pastes, uploads and prompts are checked on the computer as they happen: personal and health data, credentials, your source code, and documents your sensitivity labels mark confidential.
What have they connected AI to?
MCP servers, AI keys left in files, agents running with their permission checks switched off, local models, and AI apps granted access to Microsoft 365.
How it works
Install once. See everything. Coach first, block when you are ready.
Install
A Windows installer and a browser extension for Chrome and Edge, deployed with Intune or the tools you already use. Staff see your monitoring notice before any device enrols.
See
The dashboard lists the AI in use: which tools, by whom, on which devices, what was flagged and what people chose to do next.
Coach, then block
A rule starts by explaining itself at the moment it matters. When people know it, switch it to block. Every change your admins make is recorded.
It checks on the device.
Classification runs on each computer. Our servers receive labels, counts and hashes: that a paste held two card numbers, not the numbers. File names stay on the device too.
It is not employee monitoring.
No keystroke logging. No screenshots. A prompt is read when it is sent and a paste when it lands in an AI tool, never while someone types. Browsing outside AI tools is not recorded.
It coaches before it blocks.
People are told why, at the moment it matters, and can give a reason to go ahead. Blocking needs a coaching period first, or a reason your admin records.
What it covers today
Stated plainly, including what is not there yet.- Windows 11
- Chrome
- Edge
- ChatGPT
- Claude
- Gemini
- Microsoft Copilot
- GitHub Copilot
- 70+ more AI tools
- Desktop AI apps
- AI browser extensions
- AI command-line tools
- MCP servers
- Local models
For your DPO and works council
Bring your works council in early. We make it a short meeting.
Lamplighter is built so the privacy conversation is easy to have, because the answers are in the design rather than in a policy.
- A plain record of what it collects, and what it never does.
- A monitoring notice you write, shown before any device enrols.
- Retention you set, and an audit log of every admin action.
- Customer data stored in the EU at launch, one database per customer.
What leaves the device
Stays on the device
What people type or paste
File contents and file names
Browsing outside AI tools
Keystrokes and screens: never captured
Reaches Lamplighter
The AI tool, device and person
What kind of data it was, and how sure
Counts, sizes and hashes
A summary with no content in it
By default. An organisation can choose to keep a short redacted excerpt, and that choice is recorded in its audit log.
White paper
How Lamplighter governs AI use at work
How Lamplighter governs AI use at work without reading over anyone’s shoulder: the architecture, the privacy design and the controls.
Pricing
One plan with everything that ships. Minimum 25 devices. Prices exclude VAT and are indicative until launch.
See it on your own machines.
A 30-minute call and a live demo on Windows. Bring your security and privacy questions.