For companies and IT teams · Planned
Approved AI only, on every PC you manage.
Most companies don't want zero AI. They want the AI they've licensed and checked, and nothing else. AI Breaker for business will let you allow the assistant you've approved, block the rest, and show your auditors what was blocked, using the deployment and logging tools your IT team already runs.
For Windows 11 PCs. The personal version is in testing; the business version is planned and built on the same foundation.
| Time | What happened | Event ID |
|---|---|---|
| 08:00 | Policy applied: Level 2, 1 approved service | 401 |
| 10:14 | chat.example-ai.com connections blocked, 6 attempts | 302 |
| 11:02 | localmodel.exe blocked from starting | 300 |
| 11:40 | Local change refused: set by your organization | 403 |
Each block also points to the original Windows record, so it can be checked in Event Viewer.
AI is already on your PCs. Most of it you didn't approve.
People paste work into whichever chatbot is open in a browser tab. Local model runners install without admin rights. Office, Edge and Windows each add AI features with their own switches, and new ones arrive with updates.
- Unapproved AI services
- Company documents end up in AI services you have no contract with, no data agreement for, and no view into.
- Shadow AI
- AI engines bundled inside other apps, and local models downloaded by staff, don't show up in a web filter.
- Settings scattered everywhere
- Turning off Windows' own AI means dozens of policies across Windows, Edge, Office and inbox apps, and keeping up as Microsoft adds more.
- Audit questions
- “Which AI is allowed on our PCs, and can you show it was enforced?” is a question auditors and customers now ask.
Fine-grained control, not one big switch.
Start from a level, then allow or block single items on top of it. Everything below is set centrally and applies the same way on every PC.
- Levels plus exceptions In testing
- Pick Level 1 to 4, then allow or block individual items: one AI service, one app, one Windows AI feature. The personal version already works this way.
- Approved cloud AI Planned
- Allow the AI service your company licenses and block the other chat services, in every browser and in known AI desktop apps.
- Approved AI apps Planned
- Allowlists by publisher, by product and minimum version, or by exact file hash. Anything with an AI engine that isn't on the list stays blocked.
- Different rules per group Planned
- Target levels and approvals at groups of PCs, for example engineering versus finance. Some controls apply per device rather than per user, and the admin docs say which.
- Local AI and model files Planned
- Find AI runtimes, local model runners and model files across the fleet, and block the runtimes you haven't approved.
- Leaves your own policies alone In testing
- Where your organization already manages a setting, AI Breaker doesn't overwrite it and shows that it was set by your organization.
Managed with the tools you already run.
No new console to learn first. The business version is designed to be deployed, configured and audited through Group Policy, Intune, your RMM and your SIEM.
All of the items on this list are planned. They're being designed into the personal version's foundation now, so the business version won't need a rewrite.
- Group Policy and Intune. An ADMX template and AI Breaker's own policy key set the level, mode, options and approved exemptions. Policy always wins over local changes.
- Silent install. An MSI with documented properties for the initial level, options or a full plan file, standard exit codes, and detection and remediation scripts for Intune or your RMM.
- Events your SIEM can read. AI Breaker's own event log channel with stable, documented event IDs for level changes, every change made, blocks and policy refusals. Forward it with Windows Event Forwarding or your existing agent.
- Audit reports. Every change is written to a journal before it's made, and every block references the original Windows event, so evidence can be exported and checked independently.
- Offline volume licence. One signed licence file for the fleet. No activation call and no per-device phone-home.
- Change-controlled block list. Pin the block-list version, read its signed changelog before approving it, mirror it internally, or import it by file on air-gapped networks.
- Tamper resistance. Users, even local admins, can't lower a level or remove an approval your policy sets. An emergency undo stays possible for a local admin and is logged as a warning. Signed App Control policies (tamper lock) protect the Level 3 and 4 blocks.
- Arm64. Native support for the Arm PCs you're buying, with coverage stated per architecture.
How it works on each PC.
A small service on each PC applies the policy with controls Windows documents and supports, checks that they stay in place, and records what it did.
- Cloud AI
- Each browser's official block list, Windows Firewall rules for known AI desktop apps, and DNS rules for the rest. No traffic inspection, ever.
- Windows' own AI
- Recall, Click to Do, Copilot in Office and Edge, and AI in inbox apps, turned off with the policy settings Microsoft documents for IT admins.
- Third-party AI engines
- App Control for Business blocks local model runners and the AI engines bundled inside other apps. Policies start in audit mode, so you see what would be blocked first.
- Drift checks
- If an update or a script changes a setting back, AI Breaker notices, logs it and reports it.
- Clean undo
- Every change is recorded before it's made, so lowering a level or uninstalling returns each PC to how it was.
- Emergency recovery
- A local administrator can always undo everything if something goes wrong. Under policy, that is logged as a warning your team will see.
What it will never do.
- No traffic inspection. AI Breaker never decrypts or reads what people send, so it never sees your data.
- It isn't data-loss prevention. It blocks AI services and engines; it doesn't scan documents or stop data leaving by other routes.
- Never blocks Microsoft's servers. Copilot and Microsoft's other AI are contained with Microsoft's documented settings and App Control, the way Microsoft recommends, so licensed Microsoft services keep working where you allow them.
- Never touches Defender, at any level.
- No kernel driver, no patched system files, no undocumented tricks.
What exists today, and what's planned.
We'd rather tell you exactly where things stand than sell you a slide. Nothing on this page is for sale yet.
In testing Personal version
Being tested on Windows 11 Home and Pro.
- Levels 1 and 2
- Allow or block single items on top of a level
- Activity log with Windows event IDs, exportable
- One-click undo, panic button and a recovery command
In development Personal version
Planned for the first release of the personal version.
- Levels 3 and 4 (App Control)
- Impact scan of installed apps
- Tamper lock with signed policies
Planned Business
Built on the same foundation, after the personal version.
- Group Policy and Intune settings
- Silent MSI and detection scripts
- Event channel for your SIEM
- Approved AI allowlists and services
- Volume licence, block-list pinning
- Audit reports
We're talking to IT teams now to decide what comes first. If you manage Windows PCs, your answers shape the order.
Questions
When can we buy it?
Not yet. The personal version comes first. The business version follows on the same foundation; join the waitlist and we'll write when there's a date.
How much will it cost?
Pricing for organizations will be announced before launch. Tell us how many PCs you manage and we'll include you when we set it.
Do we need a cloud console?
No. The plan is that everything can be set with Group Policy, Intune or a registry write from your RMM, and read from the event log with your own tools.
Which editions of Windows does it work on?
Windows 11 Home and Pro. Enterprise and Education include everything Pro has, so it works there too. Group Policy isn't available on Home, so Home PCs would be configured through your RMM or scripts instead.
Can we allow our licensed Microsoft AI and block personal accounts?
Separating work and personal Microsoft accounts is done with Microsoft's own tenant controls, not by AI Breaker. AI Breaker never blocks Microsoft's servers; it can turn individual Microsoft AI features on or off with documented settings.
Will it fight our existing App Control or browser policies?
It's designed not to. Settings your organization already manages are left alone and reported, and conflicts are logged so you can see them.
Does it send data about our users anywhere?
No. Logs stay on each PC and in your own log pipeline. With a volume licence file there's no activation call; the only outbound connection is downloading signed block-list updates, which you can replace with an internal mirror or file import.
Help us build it for your fleet.
Leave your work email. We'll ask a few questions about how you manage your PCs, and write when the business version has a date.
We keep your email, plus the country and network (often your company or internet provider) you signed up from, only to tell you about the launch. Write to [email protected] and we'll delete it.