PII-Addon - What it can do and how to book it
The PII add-on is an optional premium feature that automatically detects and anonymizes personally identifiable information before it is sent to an AI model. This allows you to work with sensitive information without it ever reaching the model in plain text.
In this article, we’ll show you exactly what the add-on does, how it works, and how to enable it in your workspace.
What does the PII add-on do?
Once the add-on is enabled, every request—whether a direct chat input or an uploaded document—runs through an anonymization filter. This filter detects personally identifiable information and replaces it with consistent placeholders before the request reaches the LLM.
The finished response is then automatically “reversed” so you can use it as usual. The LLM itself never saw the real data.
What the filter recognizes
Names (people, organizations)
Email addresses and phone numbers
IBANs and credit card numbers
IP addresses and location data
Addresses (street, ZIP code, city)
List is continuously updated
How anonymization works technically
Input: You enter your prompt or upload a document.
Preprocessing in Germany: Your request first goes to our Hetzner servers in Germany—not directly to the LLM.
NER analysis: A named entity recognition algorithm scans the text in real time for personal data points.
Replacement: Detected entities are replaced with consistent placeholders:
[PERSON],[EMAIL_ADDRESS],[IBAN_CODE]…LLM Processing: Only the anonymized text is sent to the selected LLM.
Back-translation: The response is automatically populated with the actual values before you see it.
💡 Important: The actual personal data never leaves the German legal jurisdiction. The LLM sees only anonymized text.
Who benefits from this add-on?
The add-on is particularly relevant if you:
Want to use top global models or US-hosted models—while working with real customer names, IBANs, or addresses.
work in a highly regulated industry where internal guidelines are stricter than legal requirements:
Law (attorney-client privilege, client confidentiality)
Healthcare (patient data, medical confidentiality)
Finance (banking secrecy, MaRisk requirements)
Public administration with its own data protection guidelines
In practice, the following teams stand out in particular:
Legal departments that review contracts
HR teams that screen job applications
Sales & Support with direct customer contact
Compliance teams with strict requirements
Finance and accounting teams dealing with accounts and IBANs
How to activate the add-on
ℹ️ Important: Only Workspace admins can activate add-ons. If you’re a regular user, forward this article to your admin.
Step 1: Purchase the add-on
Go to Settings → Billing → Add-ons
As of May 2026, the cost is €69 per workspace per month
Once activated, the add-on will appear in your Workspace settings
Activate the add-on → Click on PII Anonymization → Activate
Step 2: Activate the add-on
Open Settings → Workspace → Add-ons.
Select PII Anonymization.
Click on Activate.
Step 3: Done
Once activated, the feature is immediately available to all users in the workspace—without anyone having to configure anything. Anonymization applies automatically to all requests.
Frequently Asked Questions from the Community
“Does the add-on affect the quality of AI responses?”
Generally not noticeably. The model works with semantically unambiguous placeholders ([PERSON], [IBAN_CODE] …), which provide it with the full context. For highly personalized tasks, there may be minimal differences—in everyday business (emails, contracts, reports), you usually won’t notice any difference.
“Does this also work with uploaded documents?”
Yes. Anonymization applies to both direct inputs in the chat and uploaded documents. You don’t have to redact anything manually.
“Can I activate the add-on per assistant or model?”
The add-on is activated at the workspace level and then applies to all requests.
“What if the filter misses something?”
No NER model is 100% perfect. The add-on is an additional technical layer of protection, not a substitute for organizational diligence (clear usage guidelines, trained staff). For maximum security, both layers complement each other.
“What happens to our data after processing?”
The actual data remains in your Workspace in Germany—encrypted with AES-256 and isolated by row-level security. As an admin, you can configure retention periods for chat content yourself (Details in the Compliance FAQ).
“How does the add-on affect our compliance situation?”
Two main benefits:
You can combine global models with personal data without this data ever reaching the US-hosted model.
You get a technically verifiable anonymization layer that you can demonstrate to internal auditors or regulatory authorities.
ℹ️ Note: innoGPT is fully GDPR-compliant even without the add-on—thanks to EU hosting, the Data Processing Agreement (DPA), Technical and Organizational Measures (TOMs), ISO 27001, and the exclusion of training data usage. The add-on is an additional layer of protection, not a prerequisite for GDPR compliance.
Do you have any questions?
Write to us—we’ll help you find the right configuration.
Related articles: