How to manage shadow AI without blocking it.
Blocking AI tools moves shadow AI out of sight. Approve fast, follow the spend, keep the inventory current. A practical guide for security and IT.
- Topic
- Shadow AI
- Product
- Novie Control
- Format
- Guide
- Reading time
- 3 min
To manage shadow AI without blocking it, make the approved path faster than the unapproved one. Publish a list of approved AI tools, answer requests for new ones in days, watch spend and identity for AI you did not approve, and keep an inventory of every AI system and agent with an owner. Blocking moves shadow AI to personal devices and accounts, where it is harder to see and impossible to govern.
What shadow AI is
Shadow AI is the use of AI tools without approval or oversight. It includes a personal chatbot account used for work, a browser extension, an AI feature switched on inside a licensed product, and an agent a team built and runs on its own credentials. It is shadow IT with a difference. A shadow tool stores data outside control. A shadow agent acts outside control.
Why it matters
IBM's Cost of a Data Breach Report 2026 found shadow AI involved in 43 percent of breaches, up from 20 percent the year before, across 602 breached organisations. The average shadow AI incident cost $5.39 million, and more than two thirds of organisations had no governance process in place to limit shadow AI.
Closer to home, use has outrun the rules. The EMA's survey of more than 300 of its member businesses, published in July 2026, found 83 percent use AI and 13 percent have an AI policy. In 18 percent, AI is used without a clear owner.
Why blocking backfires
People use unapproved AI because it helps them do their work. Block it on the corporate network and the work moves to a phone, a personal laptop or a home account. The organisation loses the one signal it had, the traffic, and keeps the risk. A blocked tool that someone needs becomes a tool used out of sight.
Blocking has a place: a tool with known data handling problems, or a class of data that must never leave a system. As a general policy it treats a demand signal as a threat.
Six steps that work
- Publish the approved list. The AI tools the business supports, the data each may handle, and who to ask about anything else.
- Answer requests fast. A request for a new tool answered in days removes most of the reason to go around the list.
- Follow the money. Card statements, expense claims and software renewals show AI tools and features nobody listed. Spend is often the first place shadow AI is visible.
- Follow identity. Service accounts and API keys with no human owner, and OAuth grants to AI tools, show what is running on the organisation's behalf.
- Keep the inventory. Every AI system and agent, with an owner, its data and its approval. Shadow AI is an inventory problem before it is a policy problem.
- Bring it in. When you find an unapproved tool that people rely on, approve it with a scope or replace it with one that is approved. Treat each find as a request that arrived late.
Agents need more than a policy
A policy tells people what not to do. An agent built last month by a team on its own credentials is already doing something, and keeps doing it whether or not anyone reads the policy. For agents, the controls are an owner, an approved scope with a review date, a budget cap and a record of what the agent did. Those sit in the path the agent's calls take, rather than in a document.
How Novie Control does it
Novie Control is live. Every dollar visible. Every budget governed. Every call optimised. Every agent governed. Every AI dollar is traced to the team, the project and the owner at the moment it is spent, so spend on AI outside the governed path shows up as a line with no owner. Every agent carries a named owner, an approval trail and a full audit trail, and caps hold before spend is committed, with a named human on every increase.
Novie OS, launching soon, makes the governed path the easy one: people describe the work in their own words, get approval, and automate it, without an IT queue.