Shadow AI is work-related use of an AI tool or feature without an organization’s required approval, awareness or oversight. It can include a personal chatbot account, an unreviewed browser extension or a connected AI agent.
How it happens
A worker may use a familiar tool because it is convenient, the approved option lacks a feature or there is no approved option. Some personal use may be permitted by an employer. Whether it is shadow AI depends on the organization’s rules and visibility, not simply the name of the product.
Why it matters
Pasting a document into a tool shares that document with the service under its applicable terms. Connecting an app or agent can grant access to files, email or other resources, depending on the permissions and settings. Neither using an unapproved tool nor connecting it proves that a breach occurred.
DNSFilter’s 2026 survey describes reports of unapproved AI connections among a sample of IT and security professionals. Our shadow AI workplace report explains the research and its limits.
What you can do
Employees should check workplace policy, avoid sharing sensitive information with unapproved tools and report accidental exposure promptly. Employers can offer a usable approved route, inventory AI applications and connections, restrict permissions and make it easy to report mistakes. If a credential was exposed, revoke or rotate it through the issuing service and follow the organization’s incident process.
See our guide to protecting a business from hackers for a wider security baseline.
