If your company has no clear rules on AI use, that doesn't mean AI isn't being used. According to Verizon's 2026 Data Breach Investigations Report, 45% of employees now regularly use AI tools on their work computers, up from just 15% a year earlier. Two thirds of them access these tools through personal accounts rather than company ones.
In other words, AI is most likely already being used in your company. The only question is whether you know who is using it, where and how.
Where the risks are
Using AI is not the problem in itself. The problem starts when it's used without rules and without anyone knowing:
- Company and client data in personal accounts. A client contract, a financial report or a piece of code pasted into a free chatbot leaves the company. You don't know where that data is stored or whether it will be used to train models. The same Verizon report found that source code was the most common type of data uploaded to external AI tools.
- Made-up facts reach clients. AI writes confidently even when it's wrong: it cites sources that don't exist, gets numbers wrong or invents terms. If nobody checks the output, the mistake ends up in a proposal or report sent to a client.
- Nobody checks the output. When AI is used in secret, there is no clear review process. A colleague who doesn't know a text was generated has no reason to look for mistakes in it.
- You don't know which tools are in use. Without that information, you can't assess either the risks or compliance. And the AI Act requires companies to take measures so that employees use AI knowingly.
Why employees hide their AI use
Most employees don't hide their AI use because they're doing something wrong. The reasons are usually simpler:
- Fear of looking lazy. If a report took half an hour to write, people may ask whether it's really "yours".
- Unclear rules. When nobody has said whether it's allowed, it feels safer to say nothing.
- Fear of being replaced. If AI does part of the job, won't it become obvious that the employee is no longer needed? Many professionals ask this, even though AI complements their work more than it replaces it.
The result is the same in every case: the employee uses AI alone, without checks and without rules.
Why bans don't work
The first reaction is often to ban AI. But a ban doesn't remove the need. An employee who saves an hour a day with AI won't stop. They'll just use it on their phone or personal laptop, and the company loses even the little visibility it had.
What matters is embracing the change instead of banning it: openly acknowledging that AI is being used and teaching people to use it safely. It's important to approve the right tools, decide what company data they can access and set clear rules. But rules only work when people understand them and have no reason to get around them.
What training changes
Many companies still don't provide training. According to a 2025 Cornerstone survey, a third of US employees whose employers encourage them to use AI receive no training for it. That leaves employees to learn on their own and to decide on their own what is safe and what isn't.
Good training changes several things:
- AI use comes into the open. Training shows that management supports AI use, so employees no longer need to hide it. People talk about which tools they use and what for, and the company finally sees the real picture.
- Clear limits appear. Employees learn which data must never be pasted into public tools and why, and how a free account differs from a company plan.
- Checking becomes a habit. Once people understand why AI makes mistakes, they start checking the output themselves.
- Results improve. Employees who know what AI can do use it more often and more wisely.
We see this in our own company too: over six months of internal AI training, mistakes with AI have become something teams discuss openly instead of hiding.
Where to start
The first step isn't a new policy. Start with a conversation: ask your teams which AI tools they use and what for. Their answers will show where the biggest risks are and what your teams are missing. From there, AI training for employees led by people who build AI systems every day will help. If you'd like to discuss what your team needs, get in touch.