Home / Glossary / AI discovery
Glossary
What is AI discovery?
AI discovery is the work of finding every AI agent, MCP server, skill, model and app that a company's employees run. A security team can only stop high-risk AI it knows about, so discovery is the step that makes blocking possible. AISeeker is AIBound's AI discovery for employee devices.
By the AIBound team · Updated 17 November 2026
Key takeaways
- AI discovery finds the AI agents, MCP servers, skills, models and apps that employees run at work.
- A security team can only stop high-risk AI that it has found.
- Much AI runs on the device, so discovery has to read the device as well as the network.
Why AI discovery matters
A rule to block a risky AI agent does nothing if nobody knows the agent is there. Employees install coding agents, MCP servers and skills in minutes, often without a ticket. Many security teams first hear about them from a tool bill, an audit or an incident.
Discovery closes that gap. It tells the security team which AI is running, who runs it and what it can reach. With those three facts, the team can stop the high-risk AI and leave the safe AI alone. A blanket ban is the usual result when those facts are missing, and bans push employees to personal accounts.
How AI discovery works
AI discovery reads the places where AI shows up and joins the results into one view. The work repeats, because new AI arrives every week.
- 1 Read the sources: employee devices, network traffic, sign-in records and SaaS admin consoles.
- 2 Identify each AI: the agent, MCP server, skill, model, app or extension.
- 3 Add the context: the device, the person who runs it and the systems it is wired into.
- 4 Pass each finding on for a decision, so high-risk AI is stopped and the rest is allowed.
AI discovery vs an AI inventory
AI discovery is the act of finding AI. An AI inventory is the list that discovery produces. Discovery keeps running, and the inventory is only as current as the last time discovery ran.
| Question | AI discovery | AI inventory |
|---|---|---|
| What is it? | The work of finding the AI in use | The list of the AI that was found |
| When does it happen? | All the time, as AI appears or changes | Each time the list is updated |
| What is it for? | Showing security what it has to act on | Recording what runs, who owns it and what was decided |
Sources
- NIST AI RMF Playbook: Govern · airc.nist.gov
Sources checked 19 November 2026
Examples
AI discovery examples
FAQ
Frequently asked questions
What is AI discovery?
AI discovery is the work of finding every AI agent, MCP server, skill, model and app that employees run at work. It shows the security team which AI is present, who runs it and what it can reach. With that view, the team can stop the high-risk AI and allow the rest.
How do you discover AI agents on employee devices?
Read the device itself. A coding agent or a local model can run on a laptop without sending traffic that a network tool would flag. A device-level check lists the agents, MCP servers, skills and models that are installed, along with the person who runs each one and the systems it connects to.
What is the difference between AI discovery and shadow AI detection?
AI discovery finds all the AI in use, approved or not. Shadow AI detection is the part of discovery that looks for AI security never approved. Discovery gives the full list. Shadow AI detection points to the items on that list that nobody reviewed and that may need a rule.
What happens after AI discovery?
Each finding needs a decision. With AIBound, AISeeker finds the AI and each finding can be allowed, asked about or denied under a policy you set. AgentHarness enforces that policy inside the agent in real time, and AIAssure records every decision as evidence for your GRC program.