Editor’s Note: This article was written by Chuck Gallagher. Gallagher is a business ethics and AI ethics keynote speaker, author and consultant. He works with executive teams and industry associations on the practical questions AI is putting in front of leadership, before those questions arrive as lawsuits, headlines or uncomfortable phone calls from customers.

Ron (OK, the name is made up) caught me after a seminar in Charlotte last spring. He had a seven-page letter from the competitor’s attorney. My guess, he hadn’t slept for nearly a week, based on the anxiety in his face. The letter stated that Ron’s company had prepared a bid using proprietary pricing that could only have come from a leaked internal document, and the evidence they cited was right there in Ron’s own submission – three sentences of the competitor’s spec language. Word for word, the competitor’s language was in his bid. Busted!

As best he could piece it together afterward, the situation was this: an estimator, trying to save himself an hour on a Friday afternoon, copied a pile of competitor product sheets into the free version of ChatGPT and asked it to compare them with his own prices and produce some persuasive content. The tool carried out precisely what had been asked of it. It also returned some text which seemed to have been left somewhere in its training data, and no one noticed it, since no one would go looking for it.

Ron said that he had read the letter three times before the real issue struck him. He had no idea what his own company had been doing. His IT policy didn’t mention AI, nor did the handbook or the NDAs he had with his vendors. Only when he finally sat his staff down and gently asked them who had been using AI for work did he get answers, which, in his words, were far more than he had wanted to hear.

That’s the problem organizations face today, and it’s not insignificant! It’s huge!

There is a term that describes what happened with Ron: “Shadow AI”, and if you run a small or mid-sized business, you likely already have some form of it in operation. Staff members use artificial intelligence to carry out their duties, yet no one in management knows which tool they chose, whose account it is being used on or what information has been fed into it. They might be using their personal Gmail account, their personal phone or the free version of whatever chatbot was mentioned at Thanksgiving. There is no IT review, and probably no written rule anywhere stating what is prohibited. It operates on the same principle I referred to in Every Choice Has a Consequence: what you do not know cannot harm you. And from the top office, all appears normal, which is what should concern you.

By the way, your people aren’t hiding anything; they never thought it was necessary to ask for permission, since that idea never occurred to them, and it feels to them just as Google did in 2005.

This becomes the company’s problem rather than the employee’s for a simple reason. If your bookkeeper copies the Q3 forecast into a free tool to help her create a chart, those figures will be stored on someone else’s server, subject to that person’s terms of service, in a data center you have never even considered. Depending on the tool and how it’s set up, the data might be used to train the next version of the model, which means your figures could end up in a product a competitor subscribes to next quarter. Your customer never agreed to that, and neither did you. And when the matter turns legal, and Ron will tell you it eventually does, the argument won’t be directed at the bookkeeper; it will be directed at the company that employs her.

What Should We Do? Three Conversations.

Ask your leadership team to gather in a room and clearly state that no one is in any trouble. Then find out who in their department is using AI, which tool they use, whether it is a personal or company account, and what kind of work is being done with it; make sure you record this information. The aim is to get a clear picture of what is actually happening within the organization, and I can assure you it will be more extensive than you imagine.

In the second conversation, you establish boundaries, and it helps to state from the beginning that not every use of AI is a problem. For example, a team member who corrects the grammar in her own meeting notes is engaging in something completely different from someone who copies a client’s lease into an AI LLM (Large Language Model – like ChatGPT). Prepare a one-page document listing everything that should not be used in a free AI tool – such as customer data, employee records, financial forecasts, contracts, pricing sheets and any other information covered by an NDA. This single page will be more useful than the 20-page template your organization eventually releases, since it reflects your business rather than a typical one.

The third conversation has a cost. Choose the tool your staff already wants, buy the paid business version and provide it to the employees who need it. With a paid account, you get data protection features which the free version lacks altogether. The information stays in your account; it isn’t sent back into the training process, it can be deleted when an employee leaves, and you will have an audit trail. It’s not an inexpensive option, but it’s a lot cheaper than discovering next spring that one of your estimators had given a competitor’s lawyer the smoking gun for a lawsuit that would cost your company a whole year.

A Final Warning

Amid the euphoria of AI and the potential it brings lies the hidden danger of ignorance. The firms affected in 2026 are generally not the ones that considered this beforehand; instead, they are those that assumed it wouldn’t happen in their building. Ask the questions. Buy the license rights. Create the AI policy. A bit of discovery, planning and forethought will be the best AI investment you’ll make. I promise.

Sign Up For Our Newsletter

Receive the latest updates on the linen, uniform and facility services industry from TRSA delivered straight to your inbox.