Shadow AI is the use of AI tools at work without the employer's knowledge or approval, usually through personal ChatGPT, Claude or Gemini accounts. It happens because AI genuinely saves time and the company has not offered an approved tool. The risk is not the AI itself. It is that client, patient or company data leaves your control, under terms nobody at your company agreed to, with no record of what was shared.
An everyday example
Picture a made-up mid-sized company, Brenner Maschinenbau. Nobody there has an approved AI tool. In one week:
- The HR manager pastes a salary table into a personal chatbot to "tidy up the formatting".
- A sales lead uploads a customer list to get a draft follow-up email.
- Someone in legal summarises a confidential supply contract with a free AI account.
Each person was trying to do good work faster. None of it was malicious. And the company has no idea any of it happened.
Shadow AI, simply explained
Shadow AI is the AI version of "shadow IT": software staff use without the IT department's approval. It usually looks like this:
- Personal accounts on consumer AI chatbots, used for work tasks.
- Browser extensions and plug-ins with AI features that read whatever is on the page.
- AI features inside other apps (note-takers, transcription tools, writing assistants) switched on without review.
It is widespread. A 2025 Bitkom survey of 604 German companies found:
| Metric | 2024 | 2025 |
|---|---|---|
| Companies reporting widespread private AI use | 4% | 8% |
| Companies reporting isolated cases | 13% | 17% |
| Companies confident no private AI is used | 37% | 29% |
| Companies that provide AI tools themselves | n/a | 26% |
Most companies either know or suspect that staff use personal AI tools at work, and only about a quarter offer an alternative. That gap is shadow AI.
The risks of shadow AI
Your data may become training data. Consumer AI plans may use conversations to improve their models unless the user opts out. Your employee probably did not.
People may read it. Providers generally reserve the right to review conversations for safety and quality. Google tells Gemini users not to enter confidential information.
You are still responsible. Under the GDPR the company is generally the controller of the personal data its staff process for work, and it must have measures in place to stop that data leaking, even through a personal account. Fines can reach €20 million or 4% of worldwide annual turnover, whichever is higher.
Confidentiality duties do not pause. Doctors, lawyers, tax advisers and others bound by professional secrecy can breach it simply by pasting a client file into a tool the client never agreed to.
AI chats are not privileged. In the US, a federal court ruled in February 2026 that a person's chats with a consumer AI tool were not protected by attorney-client privilege. Treat what goes into a chatbot as something others may one day see.
You cannot see any of it. No record of what was shared, by whom, or where it went. When a client or a regulator asks, the honest answer is "we don't know".
What shadow AI exposes that the law does not cover
The usual worry is a data breach. The bigger, quieter issue is what the provider's terms allow it to do with your content.
Data protection law protects the identifiable part: names, contact details, anything that points to a person. Recital 26 GDPR says it does not apply to anonymous information. Once a provider is allowed to reuse what you upload, it can remove the names and the rest is no longer personal data. That anonymised remainder (a salary band for a role in a city, a contract's price and liability cap, a diagnosis and its treatment) can be pooled, licensed and sold legally.
And "anonymised" is often weaker than it sounds. Researchers showed that about 87% of Americans can be identified from just ZIP code, birth date and sex.
So every personal AI account your staff use is a set of terms your company never read. We show what happens to a salary review or a board memo once reuse is allowed on the data trail.
How to stop shadow AI (without banning it)
Banning AI rarely works. Staff who rely on it find workarounds: their phone, a personal laptop, a private email address. The shadow gets darker and you lose what little visibility you had. A better plan:
- Give people an approved AI tool that is at least as good as what they use privately. This is the single biggest lever.
- Write a short AI policy: which tools are allowed, which data may go in, who to ask.
- Check the terms of every AI tool: retention, training, and whether the provider takes a licence to reuse your content.
- Sign a data processing agreement (Art. 28 GDPR) with the provider.
- Train your staff. Under the EU AI Act, organisations that use AI must take measures to ensure their staff have sufficient AI literacy.
Where Wysor fits. Wysor gives your team several AI models (GPT, Claude, Gemini and more) in one private workspace, so there is no reason to reach for a personal account. Wysor acts as a processor under Art. 28 GDPR, only on your instructions, and takes no licence to reuse your content. We never train on it and never sell it, anonymised or not. Every connected AI model provider runs with zero data retention. Data is encrypted in transit and at rest and hosted in the EU, and these commitments are in the contract, binding anyone who could ever own Wysor.
Frequently asked questions
What is shadow AI? Shadow AI is when employees use AI tools for work that the company has not approved or does not know about, most often personal chatbot accounts. It is a type of shadow IT, and it means company and client data leaves under terms nobody signed.
What are the risks of shadow AI? Confidential data can be stored by the provider, read by its staff, or used for training. The company stays responsible under the GDPR and any professional secrecy rules, without any record of what was shared.
What is an example of shadow AI? An HR manager pasting salaries into a free chatbot, a lawyer summarising a client contract with a personal AI account, or a sales team installing an AI browser extension that reads the CRM. All without the company's approval.
How do you prevent shadow AI? Offer an approved AI tool that is good enough that nobody needs a personal account, back it with a short AI policy, and train staff. Bans on their own mostly push the use further out of sight.
Is shadow AI the same as shadow IT? It is a subset. Shadow IT is any software used without IT approval. Shadow AI is the AI part of it, and the risk is higher because people paste whole documents into it.
Further reading
- The data trail. What happens to a file once someone may reuse it.
- Legal to sell. Why anonymised data can legally be sold, with sources.
- Is ChatGPT safe?. What happens to what you type into ChatGPT.
- The EU AI Act explained simply. What the AI literacy duty means for your team.
- What is sovereign AI?. Keeping control of where your AI runs and what happens to your data.
- What is generative AI? Explained simply, with examples
Questions about rolling out AI to your team? Reach us at [email protected] or through the in-app contact form.
Sources: Bitkom survey of 604 German companies, July-August 2025.
