Shadow AI is enterprise AI use that happens outside sanctioned channels: staff working in personal accounts and consumer tools with company data, invisible to security and unrecorded for compliance. Smartflow governs it at the network path rather than the endpoint, discovering the usage, applying policy inline, and leaving a record without blocking the work.
Your Employees Are Using AI Right Now. Can You See It?
Employees are pasting customer data, source code, financial projections and strategic plans into consumer AI tools through personal accounts and browser interfaces that the security team cannot see. Most organizations discover the scale of it only when they go looking.
This is not a technology adoption problem. It is a governance gap. Employees use shadow AI because it makes them faster. The answer is not to block AI. It is to govern it.
Discover what is already happening
Shadow AI is invisible at the application layer and visible at the network path, because every call to an external model leaves the perimeter. Routing AI traffic through the Smartflow control plane turns that traffic into an inventory: which models are being called, by whom, with what in the prompt, and how often. Discovery is not a separate exercise. It is the first output of putting a gateway in the path.
Govern in the browser
The network path sees every call that crosses it. It does not see what a person is about to paste until the moment the request leaves, and it cannot warn them before they do. That is the job of Smartflow Edge, which runs in the browser alongside the assistants people actually use.
On ChatGPT, Claude and Gemini, Smartflow Edge applies the signed policy at the point of the send. It can block the send, warn the person, or prepend the policy text to what they are about to submit. A prompt preview of up to 2,000 characters goes to the company proxy with the policy version and whether the send was blocked. Hosts on the block list return a block page. Calls to the model hosts are redirected through the company’s Smartflow instance using the device key, which is what puts them back under the control plane rather than outside it.
For discovery, Smartflow Edge counts visits to known AI hosts. Hostname only — no prompt content reaches that counter. It answers how much and where, not what.
What this does not do. On an unmanaged laptop the in-page block does not hold. A determined person with developer tools, or a browser without the extension, gets around it, and we will not tell you otherwise. The durable control today is the redirect through Smartflow combined with force-install on managed devices. Closing the unmanaged gap is open work.
Enforce without blocking the work
The Smartflow control plane enforces policy across AI traffic once it is visible. Block specific data categories from reaching unsanctioned tools. Redact sensitive content inline before it reaches a model rather than refusing the request outright. Apply different policy to different directory groups. Keep the record of every decision, with the policy version that produced it.
The design principle throughout is that a control which stops people working gets routed around, and a control that is routed around governs nothing. Containment that leaves the work intact is the only kind that holds.
Related reading
Enterprise shadow AI guide · DLP for AI · Information barriers · Agent governance · Agentic AI · AI bill of materials · Model allowlist and denylist
Put this in the path of your own agents.
Policy enforced inline between your agents and every model and tool they reach, with a record bound to the human who owns it.
Request a Demo Read the docs