AIBound

Home / Glossary / CASB (cloud access security broker)

Glossary

What is a CASB (cloud access security broker)?

A CASB (cloud access security broker) is a security tool that sits between a company's employees and the cloud apps they use. It shows which apps are in use and applies the company's rules to them, such as who may sign in and what data may be uploaded. A CASB reads cloud traffic and cloud accounts, so AI that runs only on the device is outside its view.

By the AIBound team · Updated 17 November 2026

Key takeaways

  • A CASB shows which cloud apps employees use and applies security rules to them.
  • It works by reading traffic to cloud apps, or by connecting to the apps through their APIs.
  • AI that runs on the device, such as a local model or a coding agent, may send little or nothing through a CASB.

Why a CASB matters

When work moved to cloud apps, company data moved outside the office network. A firewall could see that traffic went to an app. It did not show what happened inside the app. A CASB filled that gap. It made unapproved apps visible and let security set rules for sign-ins, uploads and sharing. Today it is often sold as part of a wider cloud security service.

Many CASBs now list AI websites in their app catalogues. A CASB can show who visits an AI chat site and can stop an upload to it. What it reads is traffic. A local model, or a coding agent that talks to an MCP server on the same laptop, may never send a request that passes through the CASB.

How a CASB works

A CASB works in two ways. It can sit in the path of the traffic, or it can connect to a cloud app from the side through the app's API.

  1. 1 The CASB reads web traffic logs and lists the cloud apps in use, including the ones nobody approved.
  2. 2 It rates each app for risk, and the company decides which apps are approved.
  3. 3 In the traffic path, it checks sign-ins, uploads and downloads as they happen and applies the rules.
  4. 4 Through the API, it scans the data already stored in approved apps and flags sharing mistakes.

CASB vs AI Usage Control

A CASB governs the use of cloud apps. AI Usage Control (AIUC) governs the use of AI, including the AI that runs on the employee device. The two overlap on AI websites and differ on where they look.

CASB vs AI Usage Control
Question CASB AI Usage Control (AIUC)
Where does it sit? Between employees and cloud apps, in the traffic path or on the app's API On the employee device, where the AI runs
What does it see? Traffic to cloud apps and the data stored in them The agents, MCP servers, skills and models on the device
What does it act on? Sign-ins, uploads, downloads and sharing The actions an AI agent tries, in real time

Sources

Sources checked 19 November 2026

Examples

CASB examples

FAQ

Frequently asked questions

What is a CASB (cloud access security broker)?

A CASB is a security tool that sits between employees and the cloud apps they use. It shows which apps are in use, including unapproved ones, and applies the company's rules to sign-ins, uploads and sharing. It gives security one place to manage cloud app risk across many apps.

What are the four pillars of a CASB?

A CASB is usually described by four jobs. Visibility shows which cloud apps are in use. Compliance checks that the use of those apps meets the company's rules and its legal duties. Data security controls what is uploaded and shared. Threat protection spots risky sign-ins, stolen accounts and malware.

What is the difference between a CASB and a secure web gateway?

A secure web gateway filters all web traffic. It blocks harmful sites and applies browsing rules. A CASB focuses on cloud apps and understands what happens inside them, such as a file share or a sign-in from a new place. Many vendors now sell the two together in one service.

Can a CASB detect shadow AI?

In part. A CASB reads traffic to AI websites, so it shows who uses an AI chat site and on which account. Agents, MCP servers, skills and local models run on the device, and much of that never touches the network. Endpoint tools such as AIBound read the device to find those.

Stop the risky AI your employees already run.

AIBound blocks the high-risk AI on every employee device, in real time. The 14-day trial needs no proof of concept.

Book a Demo