How to block AI apps on managed phones and tablets with your MDM

Shadow AI on mobile rarely looks like a breach. It looks like a sales rep summarizing a contract in the ChatGPT app between meetings, a support lead pasting a ticket thread into Claude to draft a reply, or an engineer photographing a whiteboard and asking an assistant to explain it. Each one is a normal workday task that moves confidential data to a third party your security review never covered.

The usual controls do not reach it. A web proxy sees the office network, not the cellular connection. App protection policies wrap the apps you push, not the assistant the person installed from the store. A written policy asks people to remember a rule at the exact moment the shortcut is most tempting. What worked on the computer, which is removing the option, is what the mobile app does on the phone.

A phone cannot run a browser extension, so the mobile app blocks the destination rather than the prompt. Once protection is on, the device cannot reach known AI destinations: chatgpt.com does not load in Safari or Chrome, and the ChatGPT app cannot connect. The same is true for Claude, Perplexity, Grok, Copilot, and the 70+ other AI tools in the catalog, and the list refreshes on its own without an app update.

Everything happens on the device. The app downloads the list of AI destinations and your policies and answers each lookup locally; it never uploads which sites or apps were used, and nothing anyone types leaves the phone. iOS shows the filter as a VPN configuration because that is the mechanism iOS provides for an app to filter, but no traffic is routed through Nullgen or any remote server.

The rollout is three items in your MDM, all generated for you on the Provisioning page of the Nullgen dashboard: the app itself, an app configuration that enrolls each device, and an optional configuration profile that turns protection on without a prompt. Here is the sequence.

One launch of the app completes enrollment. From then on the device keeps itself current, refreshing the catalog and your policies on its own, and it appears on the Devices page assigned to the person who uses it. Allow nullgen.ai and cdn.nullgen.ai on outbound HTTPS; the Deployment for IT page on the Trust Center has the details. Nothing is re-issued when the desktop agents ship: the same enrollment token covers browsers, phones, and computers.

The design goal was the same as the extension’s: remove the shortcut without surveillance and without a daily reminder that a control exists.

A block with no exit becomes a workaround. Nullgen’s answer is the same on the phone as on the computer: access requests routed to the people who should decide, with a time limit. A request for Perplexity for the afternoon reaches the approver you assigned, the approval unblocks the device the moment it is granted, and the block returns on its own when the time is up.

Nullgen blocks known AI apps and websites, and the list keeps growing. Some AI features, though, live inside a larger app the phone needs for other things, and blocking those would break search, mail, or messaging. The app leaves them alone. Your MDM can remove those apps entirely where the policy calls for it.

Two more boundaries worth stating plainly. The app protects devices your MDM manages; a personal phone that is not enrolled is out of scope, and devices under User Enrollment approve the VPN once because Apple does not deliver the profile to them. And the app filters network destinations, so an assistant built into the operating system is a job for your Restrictions payload, not for Nullgen.

The Android app will block the same catalog the same way, pushed through managed Google Play as a force-installed app with a managed configuration. It will use the enrollment token you already have, and the Provisioning page will carry its payloads when it ships. Subscribe to the newsletter and we will let you know.