# The Hidden Risk of Shadow AI in the Workplace

An employee pastes a contract into a free AI tool to get a quick summary. It does not feel like a security incident. It might be one.

It starts small. An employee pastes a contract into a public AI chatbot to get a quick summary before a meeting. Someone else uploads a customer list to have it reformatted. A developer asks a chatbot to debug a snippet of proprietary code. None of it feels like a security incident in the moment, it feels like using a tool to save twenty minutes.

"Shadow IT[↗](/shadow-it)", software employees adopt without approval, has existed for decades. Its newest form, shadow AI, is spreading faster than almost anything security teams have dealt with, mostly because a free chatbot is one browser tab away and needs no procurement process at all.

The real exposure: many free AI tools retain submitted data to improve their models unless a user actively opts out, and few people read the terms of service closely enough to know if that applies to them. Once sensitive data has gone into a third-party AI tool, the organization generally has no way to retrieve it, delete it, or even confirm how it's being used from that point forward.

Blocking access outright rarely works as cleanly as it sounds, employees who lose access at the network level often just switch to a personal device, which pushes the same risk further out of sight rather than removing it. What tends to actually work is pairing a short, plain-language acceptable-use policy with one approved, vetted AI tool that covers the same need.

If you manage a team, the cheapest diagnostic available to you today is simply asking, informally, what AI tools people already use for work. The answer is usually more revealing than any formal audit.

Source: [NIST AI RMF Generative AI Profile](https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence)