My staff are using ChatGPT without telling me. What are the risks and what should I do?

By Ranjit Bhinge, founder of Blur Studio · Updated · 6 min read

Short answer: Shadow AI means staff using AI tools such as ChatGPT (OpenAI's chat assistant) for work without the business knowing or approving it. The main risk is that customer or company information ends up in an account you do not control, under terms you have not read. Do not ban it. In one week you can ask what people use, list the tools, set a one-page data rule, provide an approved paid option, and move one real task into a reviewed workflow.

Ranjit Bhinge

You have found out, or suspect, that your team is putting work into AI tools you never approved, and you want to fix it without a fight.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

If you want to see which tasks your team is quietly doing with AI and which are worth building properly, the AI Strategy Sprint maps that in 3 days.

If you suspect it, assume it is true. Microsoft and LinkedIn's 2024 Work Trend Index found that 75% of knowledge workers use AI at work, and that 78% of AI users bring their own tools. At small and medium companies the share was 80%. Microsoft sells AI products, so treat these as survey results, not exact counts. But the direction is clear: staff are not waiting for permission, because the tools help them.

That last point shapes the whole response. Your team is not being sneaky. They found a faster way to do a task and did not know whether it was allowed. Treat it as unmet demand for a better workflow, not as misconduct, and you will get honest answers.

What the real risks are

  1. Data leaves your control. Free and personal AI accounts can have different terms from business plans. OpenAI says that by default it does not train its models on business data from ChatGPT Business, ChatGPT Enterprise or its API, unless you opt in. Personal accounts are a different case, so check the settings and terms of any account people actually use.
  2. Confidentiality promises get broken. A client contract may forbid sharing their information with third parties. Pasting it into a chat tool can be a breach even if nothing bad happens next.
  3. Wrong answers reach customers. AI tools can state invented facts in a confident tone. If nobody checks, an invented price, date or policy goes into an email.
  4. Nobody can see it. If an employee leaves, their chat history and saved prompts leave with their personal account. You also cannot tell which work was drafted by AI.
  5. Personal data rules. The US Federal Trade Commission tells businesses to collect only the personal data they need, keep it safe and dispose of it securely. Customer data in an account you cannot see or delete from does not fit that. Rules differ by country and industry, so ask a lawyer if you hold regulated data such as health or financial records.

Note what is not on the list: AI itself. The tools are fine. The risk is in the unmanaged use.

Why a ban does not work

A ban sends the use out of sight. People keep using their phones and personal accounts and stop telling you. You lose the one thing you need, which is knowing what they do. A ban also tells your best people that you would rather they were slower. A clear rule plus a good approved tool works better, because following the rule is easier than breaking it. For what is safe to paste and what is not, see whether it is safe to put company data into ChatGPT.

The five-step first-week plan

Step 1: Ask (day 1)

Send a short message to everyone, and say it plainly: "I am not in trouble-hunting mode. I want to know which AI tools you use for work and what for, so I can get you better ones and set safe rules. Nothing you tell me will be held against you." Then deliver on that promise. If the first honest answer gets someone in trouble, you will hear nothing more.

Ask three things: which tools, for which tasks, and what kind of information goes in.

Step 2: List (day 2)

Put the answers in one simple table:

Person or roleTool and account typeTaskData going inRisk
Office managerChatGPT, free personal accountDrafting quote follow-upsCustomer names and pricesMedium
Sales leadBrowser AI extensionSummarising call notesClient detailsHigh
MarketingPaid image toolSocial postsPublic material onlyLow

The example rows are made up. Check subscriptions and expense claims too, since paid tools show up there. My guide to auditing your AI tool subscriptions walks through it.

Step 3: Set the data rule (day 3)

Write one page. Three lists are enough:

  • Never put in: passwords, card and bank numbers, health details, anything under a client confidentiality clause, and personal data about customers or staff.
  • Only in the approved tool: internal drafts, notes and anonymised examples.
  • Free for any tool: public information and your own published material.

Add two more lines. First, a person checks anything AI-drafted before it goes to a customer. Second, one named person answers "can I put this in?" Also turn on multifactor authentication, meaning a second check such as a code on a phone, for every AI account, as the US Cybersecurity and Infrastructure Security Agency (CISA) advises, especially for email, social media and financial accounts. This is what people call an AI policy for a small business. Keep it to a page so it gets read.

Step 4: Provide an approved tool (day 4)

Pick one or two business plans and pay for them. A business plan gives you an admin who can manage the workspace, and clear terms about your data. Read the vendor's privacy page before you decide, using my guide to tools that do not train on your data, and see the ChatGPT for business guide for what the plans include. Then move people's work accounts over, and ask them to stop using personal accounts for work.

Check the vendor's pricing page for the current monthly price per person, and compare it with the time your team saves.

Step 5: Move one real task into a reviewed workflow (day 5)

Take the most common use on your list and make it official. Write down the steps: where the input comes from, what the saved prompt is, who checks the output, and where the result goes. Now it is a process, not a habit, and you can improve it. See human-in-the-loop AI workflows for how the check step works, and the AI implementation guide for turning it into something that runs on its own.

What to do if something sensitive already went in

  1. Find out what was pasted, where, and when. Ask the person calmly and write it down.
  2. Check the account's settings and the vendor's terms for how the data is handled, and whether it can be deleted.
  3. Delete the conversation and check the history and any saved memory in that account.
  4. If the data was client confidential or personal data, read your contract and check the law that applies to you. Ask a lawyer whether you must tell anyone. This is the point where legal advice is worth paying for.
  5. Fix the process with steps 3 and 4, so it cannot repeat by default.

Keeping the rule alive after week one

A rule written once and forgotten will fade. Put it in the new-hire checklist so every starter reads it on day one. Repeat it in a team meeting after a month, and ask what has changed in how people use the tools. Review the approved tool list every quarter, because vendors change their terms and plans. When someone asks "can I put this in?" and the answer is not obvious, add it to the one-pager so the next person does not have to ask.

Also give people a reward for speaking up. If someone finds a good use, let them show the team. If someone makes a mistake and reports it quickly, thank them. Fast reporting is worth far more to you than a perfect record that hides problems.

Who should not wait

If you handle health records, financial advice, legal files or children's data, do not wait a week. Ask a lawyer now what applies to you, and stop sensitive data going into any tool until you have the answer. The same goes for any business with strict client confidentiality terms.

How I'd approach this

I would assume it is happening and thank people for telling me. Then I would spend a week on the five steps and a small amount of money on an approved plan. The cost of doing nothing is the habit hardening into something you cannot see.

The deeper point is that every item on your list is a task someone wanted done faster. Those are your best candidates for proper workflows. Training helps too, and my guide on how to train employees to use AI has a four-week plan, while the employee training plan template keeps the records tidy. If you want help picking which hidden uses to turn into real systems first, the AI Strategy Sprint is a paid 3-day first step that finds where AI could create real new value in your business and writes down the exact steps, starting with a free 15-minute AI brainstorm call.

Questions people ask

What is shadow AI?

It is the use of AI tools for work without the business knowing about it or approving it. A common case is a staff member using a free personal ChatGPT account to draft customer emails. It is similar to shadow IT, where staff use software the company never approved.

Can I just block ChatGPT on the office network?

You can, but people will use their phones and home connections, and you will lose visibility. A block works best alongside an approved tool and a clear rule. Blocking alone moves the use out of sight.

Is it illegal for staff to put customer data into ChatGPT?

Not automatically, but it can breach client contracts or data protection rules, depending on the data, the account and where you operate. Ask a lawyer about your own situation. The FTC's guide to protecting personal information is a good starting point for US businesses.

What should a small business AI policy include?

One page: the approved tools, what data may and may not be entered, that a person checks anything sent to a customer, and who to ask. See the employee training plan template for how to roll it out.

How do I find out which AI tools my staff use?

Ask directly and promise no punishment, then check software subscriptions and expense claims. The AI subscription audit guide shows how to review them.

More in this topic: Problem solving: AI strategy and adoption (24 guides) →

Sources

Ready for the next step?

A free 15-minute AI brainstorm. We find 3 broad areas where AI could help your business.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.
See the AI Strategy Sprint

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…