prAIvicy
Use AI. Protect data.
prAIvicy makes AI usage privacy-friendly: personal data is anonymised automatically before it reaches ChatGPT, Claude & co. — and re-inserted in the response. As a ready-to-use AI assistant for your team, or as an API with one line of code.
Your prompt
Draft a payment reminder to John Smith ([email protected]). Invoice 2026-114 for €4,860, IBAN DE89 3704 0044 0532 0130 00.
Contains personal data
prAIvicy anonymises — in Germany
What the AI model sees
Draft a payment reminder to [PERSON_1] ([EMAIL_1]). Invoice 2026-114 for €4,860, IBAN [IBAN_1].
Anonymised · ready for the AI model
Every company knows this dilemma.
Teams want to work with AI — but almost every prompt contains customer names, email addresses, IBANs. And those end up in plain text with US providers.
Every prompt
containing personal data sent to a US provider is a data transfer under the GDPR (Art. 44 ff.) — even if it is "just a quick summary".
up to €20M
or 4% of annual revenue is what violations can cost (Art. 83 GDPR). Plus reputational damage that no number captures.
Bans do not help
AI bans only create shadow usage through private accounts. The sustainable answer is a technical solution that makes safe usage possible.
One shield, two ways in.
Whether your team wants to start right away or your developers need to secure an existing integration — everything runs through the same anonymisation layer in Germany.
The AI assistant for your team
Log in, pick your profession, get to work. Chat with documents, audio and images — every message is anonymised before it reaches the AI model.
- Ready to use, no setup
- Profession-specific assistance — from lawyer to pharmacist
- Playground: watch the anonymisation live
The API for your integrations
OpenAI/Anthropic-compatible endpoint: swap the base URL, done. Your existing applications send anonymised instead of raw data to the models — without changing any code.
- One line of configuration
- BYOK — your provider keys stay yours
- Also detects secrets: API keys, credentials, private keys
An assistant that knows your profession.
Not a blank chatbot: the built-in assistant adapts its working style, quick actions and knowledge sources to your profession — built especially for professions working with confidential client and patient data.
Eight profession profiles at launch
Profession profile: Lawyer · Employment law
Quick actions
Answers cite from the legal database — sourced, not hallucinated:
§ 622 BGB§ 623 BGBSource & date includedCitable legal database
German statutes (BGB, HGB, ZPO, StGB, GG and more) right in the chat. Sections are cited and linked from the database — with built-in protection against invented references.
Documents, audio & images
Upload and search PDF and Office files, transcribe dictations, analyse photos — right in the chat, no extra tools.
Drafts on a canvas
Contracts, pleadings, letters and reports take shape in an editable document right next to the chat — versioned and exportable.
Teams, matters & budgets
Organise mandates and projects as matters, share conversations with your team, keep costs per member in view.
How the anonymisation works
Assistant or API — every request passes through the same four steps before it reaches an AI model.
- baseURL: "https://api.openai.com/v1"
+ baseURL: "https://api.praivicy.eu/v1"The request goes to prAIvicy
Automatically from the assistant — or from your own integration pointing to prAIvicy instead of the AI provider. One line of configuration.
prAIvicy detects & anonymises
Names, email addresses, IBANs, tax IDs and dozens more data types are detected automatically and replaced with placeholders — on servers in Germany.
Anonymised data reaches the model
The AI model works with [PERSON_1] instead of real names — detected personal data never leaves Germany, while response quality stays practically unchanged.
The response comes back complete
prAIvicy re-inserts the original data into the response. The whole process is invisible to your application and your users.
Built for the real thing: actual business data.
A look at what is already built. The full feature list arrives with the beta launch.
German-first detection
IBAN, tax ID, VAT ID, licence plates, addresses — 30+ data types, optimised for German formats and 17 European languages in total.
More than just text
PDF and Office documents, spreadsheets, audio and images are anonymised too — including face and signature redaction.
One endpoint, every model
OpenAI, Anthropic, Mistral, Google and more through a single OpenAI-compatible API. Switch models without changing code.
Your keys stay yours
Bring your own key: you use your own provider accounts. prAIvicy only ever stores keys encrypted — never in plain text.
EU hosting, no PII storage
Processing in German data centres. The mapping from original data to placeholders lives only for the duration of a request.
Provable, not just promised
Dashboard with audit log: which data types were anonymised, and when? Compliance you can show your data protection officer.
Reserve your beta seat.
We invite the waitlist in waves — early joiners shape what gets built.
- 3 months free for beta participants
- Direct line to the engineering team
- Your feature requests get prioritised
- Limited seats per wave
Frequently asked questions
A privacy proxy for AI APIs: your requests pass through prAIvicy, personal data is detected and anonymised automatically before it reaches the AI model. The response comes back and the original data is re-inserted — your code never notices.
No. prAIvicy is API-compatible with OpenAI and Anthropic. Usually all it takes is swapping the base URL of your existing integration — one line of configuration. And if you have no integration at all, you simply use the built-in AI assistant in the browser.
Eight profession profiles at launch: lawyers, tax advisors, doctors, psychotherapists, auditors, notaries, architects and pharmacists — professions bound by statutory confidentiality benefit most from automatic anonymisation. But any company can use prAIvicy; the assistant also works without a profession profile.
From a built-in database of German statutes (BGB, HGB, ZPO, StGB, GG and more). The assistant cites exclusively from it and labels what is sourced versus its own assessment — instead of inventing references. This is no substitute for legal advice, but it makes answers verifiable.
OpenAI, Anthropic, Mistral, Google and more — through a single endpoint. You use your own API keys (bring your own key); prAIvicy only ever stores them encrypted.
Anonymisation runs entirely on servers in Germany before anything leaves the EU. The mapping between original data and placeholders exists only for the duration of each request and is never stored permanently.
The beta rolls out in waves — the waitlist gets invited first. Beta participants use prAIvicy free for 3 months, get a direct line to the engineering team and a beta agreement including a data processing agreement (DPA). After that, plans starting at €49 per month are planned.
prAIvicy is a product of BlockBuilder UG, Munich — built by the same team that helps mid-sized companies adopt AI in a GDPR-compliant way. The product grew out of exactly those projects.
No. prAIvicy is a technical measure in the sense of privacy by design (Art. 25 GDPR) — a strong building block of your compliance, not a substitute for the legal assessment of your processing activities.