FollowAgents AI AGENTS DIRECTORY
Home / AI Agent Glossary / Shadow AI
Evaluation & Safety

Shadow AI

Also called: shadow AI agents · unsanctioned AI

Shadow AI is the use of AI applications, models, or agents inside an organization without the approval, oversight, or visibility of IT and security teams.

Shadow AI is the AI-era version of shadow IT. An employee pastes customer data into a personal chatbot account, installs a coding agent on a work laptop, or connects an unapproved MCP server to company systems, all to get work done faster, and none of it is known to the security team.

The risks are concrete: sensitive data leaving the organization, credentials and tokens granted to tools no one reviewed, agents acting with a person's permissions, and compliance obligations that assume you know where data goes. Agents raise the stakes over plain chatbots because they can take actions, not just receive text.

Bans are hard to enforce and often push usage further underground, so common responses combine an approved tool list, clear policy, visibility, and sanctioned alternatives.

Example

A developer connects a community-built mcp-server to their work repository with a personal access token to speed up code review. The server was never vetted, so the organization cannot tell what data it can read, and the token appears in no inventory.

How it differs

Shadow AI vs. prompt-injection: shadow AI is a governance gap (unknown, unapproved tools in use). Prompt injection is an attack technique against a specific agent. Shadow AI makes injection and leakage harder to detect because nobody is watching those tools.

Common misconceptions

Often assumed: Shadow AI only involves people using chatbots on the side.
Actually: It increasingly includes agents, browser extensions, and MCP connections that hold credentials and can act, which are riskier than pasted prompts.

FAQ

What is shadow AI?
Employees using AI tools, models, or agents without the organization's approval or visibility, so security and compliance teams cannot see or govern them.
Why is shadow AI risky?
It can leak sensitive data, spread unreviewed credentials and permissions, and create compliance gaps, and agents can also take actions rather than only read data.
How do organizations manage shadow AI?
Typically with an approved-tools list and clear policy, discovery and monitoring of AI usage, scoped credentials for agents, and sanctioned alternatives so people do not need to go around the rules.

Last checked: 2026-09-20

Related terms

Agent Identity

Agent identity is the practice of treating an autonomous agent as its own principal in an access-management system — with its own credentials, narrowly scoped and time-limited permissions, a delegation chain back to an authorizing human, and its own audit trail.

AI Agent Safety

Agent safety is the practice of anticipating and limiting the harm an AI agent could cause once it can take real actions in the world, not just generate text.

OWASP Agentic AI Top 10

The OWASP Top 10 for Agentic Applications is a community-produced list from the OWASP GenAI Security Project that catalogues the ten most significant security risks specific to AI agent systems, from goal hijacking to rogue agents.

Agent Observability

Agent observability is the practice of capturing enough visibility into an agent's step-by-step trajectory — which tools it called, in what order, with what results — to understand and debug why it behaved the way it did.