The filter between your team and AI — personal data doesn't get through
zeropii is a proxy filter that sits in front of every AI model. It intercepts Dutch BSN, IBAN, KvK numbers and names in prompts, masks them before they reach ChatGPT, Claude or Gemini, and restores them in the answer. You run zeropii at your own organisation, so personal data stays inside.
Built for organisations that handle sensitive data: legal, healthcare, finance, HR and government.
One filter, in front of every major AI model
One pasted BSN in ChatGPT is a data breach.
Your team already uses AI, with client names, case files and numbers in it. Every prompt with personal data that goes to an external model is a notifiable risk under the GDPR. AI isn't the problem. The problem is that nothing sits between your team and the model.
of employees who use AI share sensitive company information with AI tools, without their employer's knowledge.
Source: National Cybersecurity Alliance & CybSafe, Oh, Behave! 2024–2025
data breaches reported to the Dutch Data Protection Authority in 2024, nearly 50% more than the year before. Human error is the leading cause.
Source: Dutch Data Protection Authority (AP), Data Breach Report 2024
zeropii is that checkpoint: a filter that strips personal data before it leaves.
The filter sits in between
Every prompt goes through zeropii first. Personal data out, then to the model, and the full answer back.
Employee types
"Send invoice to Jan de Vries, BSN 123456782"
zeropii filters
"Send invoice to ████████, BSN ████████"
AI answers safely
The answer comes back through zeropii with the real names. The model never saw them.
Three pillars
Detect, mask and make it provable. That's what an AI firewall should do.
Detection
Recognises Dutch personal data — BSN, IBAN, KvK, licence plates, names, addresses — plus your own organisation-specific patterns.
Masking
Replaces personal data with surrogates before the LLM call and restores them in the answer.
Provability
Dashboard with an audit trail: see and export what was blocked, for your own accountability.
The AI firewall for Dutch organisations
Four things that genuinely matter.
Recognises what matters
BSN, IBAN, KvK, licence plates, names, addresses — plus your own patterns, tuned to the Dutch context.
Sits in front of any AI model
Claude, ChatGPT, Gemini or your own internal model. zeropii is the filter in between.
Insight & control
Dashboard with an exportable audit trail — see what was blocked and record it.
Deployed at your place
zeropii runs as a proxy at or for your organisation, so personal data stays inside.
Protection tailored to you
Standard protection for all common personal data, plus your own rules.
Standard
Per organisation
Where does zeropii run?
The proxy runs where you want it. We're honest about what happens where, no small print.
At your organisation
zeropii runs as a proxy on your own environment or a dedicated VPS. Personal data never leaves your organisation; only the cleaned prompt goes to the AI model.
Fully internal
Combine zeropii with a local AI model, and even the cleaned prompt stays within your own infrastructure.
The online demo runs on our environment to show the concept; in production zeropii runs at your place.
With or without a filter
Who it's for
For organisations that want to use AI without giving up personal data: legal, healthcare, finance, HR and the public sector.
Proxy rollout at your organisation, on your own environment or a VPS, on request.
What is zeropii?
zeropii is an AI firewall for Dutch organisations: a proxy filter that sits between employees and AI models. It intercepts personal data such as BSN, IBAN, KvK numbers and names in prompts before they reach models like ChatGPT, Claude or Gemini, and restores them in the answer. Deployed at the organisation, personal data stays inside. zeropii is made by Artificial Studio.
Frequently asked questions
See the filter in action
Type a text with personal data and see what zeropii detects and masks before it goes to an AI model.