Verho
Aarne
Co-founder
Aarne is the fearless inventor at Verho. He brings over a decade of experience in technology, research, leadership and operations from a global consultancy, including pioneering AI-driven solutions.
About Verho
Concerned about sensitive information ending up with American or Chinese AI giants? Verho lets you keep using the most capable AI models: it finds personal data in your text and hides it before your text reaches the AI model.
-
Accurate detection
Verho is very good at detecting personal data.
Share of names found
- 98.6 %Finnish
- 97.8 %Swedish
- 100.0 %English
-
Built for European languages
Our machine-learning model has been further trained on the major European languages to find personal data. It recognises names in their inflected forms and, where the answer needs it, tells the AI service whether a person is a woman or a man, so that pronouns come out right.
-
Easy-to-use browser extension
Verho works right in Claude and ChatGPT, in Chrome and Edge. It marks what it finds as you type, you choose what to protect, and the real values come back in the answer.
How it works
Verho hides the original sensitive information, such as personal data, from the AI service before your text is sent to it.
- Write or paste your text into Claude or ChatGPT as usual.
- The extension and our secure machine-learning model find personal data and other sensitive information in the text.
- You decide what is protected: add or remove items by hand.
- Verho replaces the sensitive information with placeholders. You then send the protected prompt yourself.
- When the answer arrives, Verho puts the original values back into it, in your own browser.
Verho works in Claude and ChatGPT. Microsoft Copilot, DeepSeek and Qwen will become supported in late 2026.
Supported languages
English
German
French
Spanish
Polish
Italian
Dutch
Swedish
Danish
Norwegian
Finnish
Information security
Verho works right in your browser, as an extension. Your browser is the only place where your sensitive information is ever stored.
To identify personal data in your prompt fast and reliably, Verho uses our own fine-tuned large machine-learning model, which runs on a secure UpCloud server in Helsinki, in our own protected environment. The server stores nothing you write: it only finds the sensitive parts of the text and tells your browser where they are. All traffic is strongly encrypted.
The Verho server is in Helsinki, in the European Union.
Pricing
| Light For individuals who want to protect their prompts. | Pro For power users and companies; supports masking in files. | |
|---|---|---|
| Finds and protects personal data as you type | Included | Included |
| The original values stay in your browser and come back in the response | Included | Included |
| Protects personal data in files you send | Not included | Included |
| Puts the original values back into files the AI creates | Not included | Included |
| Firm-level invoicing | Not included | Included |
| A monthly report for the firm | Not included | Included |
| Price | Free | €10per seat per month + VAT |
| Get Verho | Coming soon |
Enterprise
For larger organisations: price and terms are agreed separately. The model can also run on your own servers.
Call or email us: +358 46 804 7746 · aarne@tekoalytehdas.fi
The team behind Verho
Verho is a Finnish product, built by an experienced team with a background in business and operations. We have seen organisations fall back on weaker AI models, or avoid AI altogether, because of data protection concerns. We are building Verho to enable businesses and public sector to use the best tools without worrying about data protection.
Tuomas
Co-founder
Tuomas is the pragmatic builder at Verho. Over the past decade, he has worked across consulting, corporations and growth firms, launching new products and building businesses from the ground up.
Frequently asked questions
Does Verho see or store what I write?
Your text goes to our server in Helsinki to be checked for personal data, but it is not stored there. The original personal data is kept and put back only in your own browser. The one exception: if you give feedback and choose to attach the conversation, we keep the protected texts you attach, with the personal data replaced by placeholders, for 90 days.
Is Verho secure?
Verho is built to store as little as possible. The text it checks travels encrypted (TLS) to our server in Helsinki, is checked in memory and forgotten at once: it is not written to a database, logs or backups. The original values stay in your own browser, encrypted, and the AI service only ever sees placeholders. There is no password to leak: you sign in with Microsoft, Google or a one-time code sent to your email. The details are in our privacy policy.
What kind of information does Verho find?
Names, in any inflected form, personal identity codes, addresses, health information, salary information, bank account numbers, organisations, email addresses, phone numbers, passwords and keys, business IDs and case reference numbers. You can mark anything else by hand. For Enterprise customers we can add other kinds of information by agreement.
Does Verho find everything?
No tool finds everything. Before you send, you see what Verho found, and you can add anything it missed by hand. We are working hard to improve the detection accuracy by training our model (but never with your data).
Which AI services and browsers does Verho work with?
Verho works today in Claude and ChatGPT in Chrome, and in Edge through the Chrome Web Store. Microsoft Copilot, DeepSeek and Qwen will become supported in late 2026.
Is there a version for mobile phones?
Not yet, but we are exploring how to bring Verho to mobile devices in the near future.
Does Verho work outside the EU?
Yes. Verho works wherever you use Chrome or Edge on a computer, and wherever you are, the text it checks is processed on our server in Helsinki, inside the EU. It recognises personal data best in the European languages listed under How it works.
Does my AI service still understand the text when the data is replaced?
The AI service sees placeholders such as ⟦PERSON_1⟧ and writes around them. Verho also gives the AI instructions on how to handle them. Verho then puts the original values back into the answer, correctly inflected for the sentence.
Why doesn't Verho run entirely on my own computer?
Recognising personal data reliably, inflected names included, takes a machine-learning model that is too large to run quickly in a browser on an ordinary laptop. Thus, it runs on our server in Helsinki, which checks your text in memory and forgets it at once, and we can improve it without you installing anything. For organisations that need it, Enterprise can run the model on their own servers.
How does Verho relate to the GDPR?
Verho helps keep personal data out of AI services, but on its own it does not meet the GDPR's requirements for you. Your firm remains the controller of its data, and for firms we offer a data processing agreement.
Get Verho
Install Verho from the Chrome Web Store, then sign in from the extension with Microsoft, Google or your email address. Your first sign-in creates your account, so you can also start here on the website.