How do you find and govern shadow AI?
Margareta Petrovic, Chief AI Officer at DigitalXForce, wrote this page.
You find shadow AI by comparing the AI actually in use with the AI that was approved. Look in single sign-on and web gateway logs, expense and procurement records, cloud accounts and code, and the settings of software you already run. Then register each find with an owner, or switch it off.
Shadow AI is any AI tool, feature, model or agent used for company work without being approved and recorded. It is common. ISACA’s 2026 AI Pulse Poll found that 90% of respondents believe employees use AI at their organization, while only 38% of organizations have what ISACA calls a formal, comprehensive AI policy and 25% have none.
Shadow AI is a gap in AI governance. AI governance is the set of policies, roles, controls and evidence an organization uses to decide which AI systems it runs, how they may be used and whether their controls keep working.
The NIST AI Risk Management Framework asks for mechanisms to inventory AI systems, resourced according to the organization’s risk priorities (GOVERN 1.6), and for policies that address AI risks from third parties (GOVERN 6.1). An inventory that lists only what teams chose to declare does not meet the first of these.
Where does shadow AI hide?
Shadow AI sits in six places, and each needs a different way of looking.
| Where it hides | What you find there | How to look |
|---|---|---|
| Public AI tools on personal accounts | Employees paste work into chat assistants they signed up for themselves. | Compare single sign-on and web gateway logs with the approved list each month, and read expense claims for AI subscriptions. |
| AI features inside approved software | A vendor switches on an assistant inside a tool the organization already licenses. | Read vendor release notes and admin consoles, and ask about AI in supplier questionnaires and at renewal. |
| Models and AI services in cloud accounts | Teams call hosted model APIs or run AI services under a cloud account. | Inventory the AI services enabled in each cloud account, and look for model provider keys in code and secret stores. |
| AI in code, pipelines and containers | Developers build models, retrieval pipelines and agents into applications. | Scan repositories, CI/CD pipelines and container images for model files, AI libraries and vector stores. |
| Agents with service credentials | An automation or agent acts through a service account or an API key. | Review non-human identities and the permissions each one holds, and log what each agent does. |
| AI at suppliers | A supplier processes the organization’s data with AI. | Ask about it in due diligence, and write the answer into the contract. |
What should you do with each find?
- Register it with an owner, a purpose, the data it touches and the account type.
- Decide whether it may continue, and switch off what may not.
- Classify its risk by use case, data sensitivity and autonomy.
- Turn the policy statements that apply to it into controls with tests.
- Record any exception as a named person’s decision with a reason and an expiry date.
- Look again on a schedule, since new tools and new features appear between reviews.
How does DigitalXForce find and govern AI?
TRiSCM™, or Trust, Risk, Security and Compliance Management, is a category of enterprise software defined by DigitalXForce that converges automated governance, risk and compliance with security posture management into a single real-time system, replacing periodic control validation with continuous control assurance. Its AI-specific module, AI TRiSCM and AI Risk Governance, builds the AI inventory by discovering AI where it runs. It discovers AI assets across cloud, code, pipelines, containers, model endpoints and RAG stores, assesses LLMs, copilots, agents and models, and maps them to the NIST AI RMF, ISO/IEC 42001, the EU AI Act, the OWASP LLM Top 10 and MITRE ATLAS. DigitalXForce’s launch release for the Enterprise TRiSCM platform lists shadow AI detection among the areas AI TRiSCM covers.
The Digital Trust Score is DigitalXForce’s composite score from 300 to 850, computed continuously from live control evidence across seven sub-postures: security, compliance, audit, resilience, third-party, AI and risk. Its AI sub-posture covers the AI inventory, model and use-case risk, data sensitivity exposure and human oversight.
The 15 DigitalXForce modules share one data layer, so a failed control result reaches the compliance view, the posture view and the risk register at the same time. The suppliers that provide AI models are vendors like any other, and AI-Powered Third-Party Risk Management covers their onboarding, questionnaires and document analysis.
An AI control that was approved still has to keep working. Continuous Control Assurance (CCA) uses evidence, monitoring and validation to determine whether controls continue to operate as expected. Continuous Control Monitoring (CCM) monitors conditions, evidence and signals associated with controls. CCM is a capability within CCA.
What does a person decide, and what does the platform automate?
The platform automates the looking. AI TRiSCM discovers AI assets in the systems the platform connects to and assesses each one.
People decide what the organization allows. A person should approve each AI system before it goes into production, decide what happens to each find and accept any exception with a reason and an expiry date, and the organization writes the policy the inventory is tested against.
What are the limitations?
- Discovery reads the systems an organization connects. AI used on a personal account from a personal device leaves little or no trace in company systems, so a written policy, training and a regular log review are still needed.
- An inventory is not governance. Each find still needs an owner, a decision and, if it stays, controls with tests.
- New AI features arrive with ordinary software updates, so an inventory that was complete last month may not be complete today.
- DigitalXForce is not the answer for every organization. One whose only AI question is which public chat tools employees may use needs a policy, approved business accounts and a monthly log review more than a governance module, and DigitalXForce fits an organization whose AI is spreading faster than its register.
Which DigitalXForce modules are involved?
- AI TRiSCM and AI Risk Governance is the module that discovers and assesses AI systems.
- AI-Powered Third-Party Risk Management covers the suppliers that provide AI like every other supplier.
- Attack Surface Manager discovers and inventories assets across nine asset classes, IT and OT, without agents and through APIs.
- The page on Continuous Control Assurance explains how a control result is validated before anyone relies on it.
- The pillar page on AI governance explains how to run it continuously.
- The page on turning an AI policy into testable controls covers what happens after a find is registered.
- The article on a workplace policy for public AI tools covers which public tools employees may use.
- How DigitalXForce works lists the other pages in the knowledge center.
Which integrations and frameworks does it draw on?
AI TRiSCM discovers AI assets across cloud, code, pipelines, containers, model endpoints and RAG stores, and the platform connects to the tools an organization already runs through 250+ technology integrations. The integrations page has the current list.
The inventory and its assessments map to the NIST AI RMF, ISO/IEC 42001, the EU AI Act, the OWASP LLM Top 10 and MITRE ATLAS, and the frameworks page shows the most requested of the 50+ compliance frameworks DigitalXForce maps, with the full list available on request.
Questions about shadow AI
What is shadow AI?
Shadow AI is any AI tool, feature, model or agent used for company work without being approved and recorded. It includes public chat assistants on personal accounts, AI features a vendor switched on, and models that teams built or called without review.
How common is shadow AI?
It is widespread. In ISACA’s 2026 AI Pulse Poll, 90% of respondents believed employees use AI at their organization, while only 38% of organizations had what ISACA calls a formal, comprehensive AI policy.
How do organizations build an enterprise AI inventory?
Start from discovery rather than declarations. Read cloud accounts, code, pipelines and the settings of software already licensed, add what procurement and expense records show, and register each system with an owner, a purpose and the data it touches.
What should we do when we find an unapproved AI tool?
Register it with an owner and a purpose, decide whether it may continue, and switch it off if not. If it stays, classify its risk and put the policy statements that apply to it under test.
Can DigitalXForce find AI we do not know about?
It can find AI in the systems the platform connects to. AI TRiSCM discovers AI assets across cloud, code, pipelines, containers, model endpoints and RAG stores, so the inventory does not depend on what teams chose to declare.
Can any tool see employees using AI on personal accounts?
Discovery works on the systems an organization connects, so AI used on a personal account from a personal device leaves little trace for any tool to find. A written policy, approved business accounts and a regular log review cover that gap.
Which framework asks for an AI inventory?
The NIST AI Risk Management Framework does. Its GOVERN 1.6 outcome asks for mechanisms to inventory AI systems, resourced according to the organization’s risk priorities.
Sources
- ISACA released its 2026 AI Pulse Poll of more than 3,400 digital trust professionals on 5 May 2026, at https://www.isaca.org/about-us/newsroom/press-releases/2026/ai-use-accelerates-while-governance-and-roi-lag-says-new-isaca-research, read on 25 September 2026 and again on 26 September 2026.
- NIST released the Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, on 26 January 2023, at https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf. GOVERN 1.6 and GOVERN 6.1 were read in the published PDF on 25 September 2026 and again on 26 September 2026.
- DigitalXForce describes AI TRiSCM on its AI TRiSCM and AI Risk Governance page, in its launch release for the Enterprise TRiSCM platform and in its glossary, all read on 26 September 2026.
Do this once instead of every audit.
Every step above can be done by hand. DigitalXForce does them continuously, maps the result to 50+ frameworks once, and keeps the evidence current between audits. A 30 minute walkthrough on your own framework set shows what that removes from your calendar.



