Law Firms & Legal Departments
Use AI for research, contract review and legal briefs – without exposing client data.
- Contract analysis with ChatGPT
- Case law research
- Legal brief drafts
GDPR • AI Act • Made in Germany
Use AI –
Protect Data.
KI-Shield automatically pseudonymizes personal data before it is sent to ChatGPT, Claude or Gemini – as an additional technical safeguard in accordance with GDPR Art. 25 and Art. 32. For law firms, medical practices and enterprises.
BYOK Principle: Your own API key – full cost control.
No credit card required • 50 requests/month free
ki-shield ~ proxy
# Send client email to ChatGPT
$ ki-shield --scan "Review request"
# Sensitive data detected & filtered:
Client: Dr. James Mitchell → [PERSON_001]
E-Mail: mitchell@lawfirm.com → [EMAIL_001]
IBAN: DE89 3704 0044 ... → [IBAN_001]
✓ Request pseudonymized and forwarded to AI
Live Demo
Try It Yourself – in Real Time
Type on the left – see on the right what the AI receives instead. All personal data is automatically replaced with placeholders. No login required.
Your Input
Select a word to manually tag it as PII
What the AI Sees
Detected PII — click to remove
Tag as PII
Ø 30ms latency — real-time detection of 42 PII categories
Tip: Select a word in the input field & add it as PII • 46 recognizers • NER + Regex + Keyword
AES-256-GCM Encrypted
German Servers (Hetzner)
GDPR Art. 25/32 & AI Act
< 50ms Latenz
BYOK – Your Own API Key
Zero-Knowledge Storage + Browser ZK Mode
42
PII Categories Detected
10+
AI Providers Supported
<50ms
Average Latency
100%
German Servers (EU)
The Problem
AI Is Powerful – but a Data Protection Risk
Anyone who sends personal data to ChatGPT, Claude or Gemini without safeguards risks GDPR violations. The fines are severe.
Law Firms & Attorneys
Client names, case numbers, contract details – all sent unencrypted to OpenAI. A violation of attorney-client privilege.
Doctors & Hospitals
Patient names, diagnoses, medical reports – medical confidentiality protects this data. Using AI without pseudonymization can be a criminal offense.
Enterprises & Corporations
Customer data, financial figures, personnel records – GDPR fines up to €20 million or 4% of annual revenue.
€4.15B
GDPR fines since 2018 (EU-wide)
68% of companies already use AI tools – but only 23% have a data protection strategy for it.
Source: Bitkom Digital Office Index 2025
Source: Bitkom Digital Office Index 2025
Target Audiences
For Everyone Who Refuses to Compromise on Confidentiality
KI-Shield was built for industries where data protection is not optional – it is mandatory.
Doctors, Practices & Hospitals
AI-assisted documentation, medical letters and findings – without sending patient data to US servers.
- Draft medical letters with AI
- Finding summaries
- Medical documentation
Tax Advisors & Auditors
Financial analyses, tax assessments and expert opinions with AI – client financials stay protected.
- Review tax returns
- Balance sheet analysis
- Expert opinion drafts
Enterprises & HR
AI in HR, sales and customer service – without compromising employee or customer data.
- Screen applications
- Customer emails
- Internal reports
Data Protection Officers
A tool you can confidently approve for AI usage across your organization.
- Compliance reports
- Tamper-proof audit log
- Technical measures documentation
Software Companies
Integrate AI into your software without transmitting user data to third parties.
- PII-Redaction REST API NEW
- npm SDK + Swagger-Docs
- On-premise available
How It Works
Protected AI Usage in 3 Steps
No code changes, no SDK. Just add your API key – done.
01
Add Your API Key
Sign up for free and add your own API key (OpenAI, Anthropic, Google & more). Your app sends requests to KI-Shield instead of directly to the provider.
02
Automatic Data Pseudonymization
Intelligent PII detection identifies names, emails, IBANs, diagnoses, case numbers and 37 more data types — 42 categories in total. Everything is pseudonymized with secure placeholders.
03
Re-hydrate the Response
The AI response comes back, placeholders are replaced with the original data. Nothing changes for you or your team.
Features
Enterprise Security, Easy to Use
Built for the strictest data protection requirements – usable without an IT department.
42 PII Categories Detected
From names and IBANs to diagnoses and case numbers to genetic data, union membership and criminal records – incl. GDPR Art. 9 & 10.
Automatic Re-Hydration
Pseudonyms in AI responses are automatically restored. Your team's workflow remains unchanged.
Quantum-Safe Audit Chain
Hybrid Ed25519 + ML-DSA-65 signatures — classically and post-quantum secured. Exportable as CSV/JSON. Perfect for data protection audits.
BYOK – Your Own API Key
Bring Your Own Key: Use your own API key for OpenAI, Anthropic, Google & more. Full cost control, no markups.
Real-Time Streaming (SSE)
Real-time streaming in Pro and Business plans. Pseudonymization happens in milliseconds – no noticeable difference.
Industry-Specific Rules
Custom filter rules for your industry: medical terminology, legal case numbers, financial data – individually configurable.
42 Data Categories
The Most Comprehensive PII Detection
42 data categories – from names and IBANs to genetic data and union membership. Everything the GDPR protects, we detect.
Core Detection (16)
Person
Email
Phone
IBAN
Tax ID
Date of Birth
Date
License Plate
Social Security
Address
Health Data
Location
Organization
Credit Card
ID/Passport
Nat./Rel./Pol.
Technical IDs (12)
IP Address
BIC/SWIFT
Vehicle ID (VIN)
GPS Coordinates
Driver's License
Health Ins. No.
Commercial Register
Case Number
MAC Address
IMEI Number
Secret/API-Key
Diagnosis Code (ICD/OPS)
GDPR Art. 9 (7)
Special categories – processing prohibited
Genetic Data
Biometrics
Ethnic Origin
Political Opinion
Religion
Union Membership
Sexual Orientation
Life Areas (7)
Criminal Law (Art. 10)
Children's Data
Financial Data
Employment
Education
Social Benefits
Insurance
Compatibility
Works with All AI Providers
One Shield, all models. Switch AI providers without rebuilding your compliance setup.
OpenAI / ChatGPT
Anthropic / Claude
Google Gemini
Mistral AI
Meta Llama
Cohere
Groq
DeepSeek
Perplexity
OpenAI-compatible APIs
For Developers
NEW
PII Detection as REST API — for Your Software
The same engine that powers our chat proxy, as an API for developers. Detect, redact and pseudonymize personal data programmatically — with a signed compliance certificate for every request.
4 Endpoints, 1 API
/v1/detect, /v1/redact, /v1/pseudonymize, /v1/redact/zk — 42 PII categories, optimized for German and English texts.
Hybrid Compliance Certificate (Ed25519 + ML-DSA-65)
Every request receives a cryptographically signed certificate with a hybrid signature (classical + post-quantum) — verifiable, exportable, audit-proof.
Split/Zero-Knowledge Mode
Optional: Only hashes are sent to the server. Original data stays with you.
SDK & Swagger Docs
npm install ki-shield-sdk — TypeScript SDK + interactive API documentation at /api-docs.
curl — PII Detection API
$ curl -X POST https://ki-shield.de/v1/detect \
-H "X-API-Key: ks_live_xxx" \
-H "Content-Type: application/json" \
-d '{"text": "Max Mustermann, IBAN DE89..."}'
# Response:
{
"entities": [
{ "type": "PERSON",
"text": "Max Mustermann",
"score": 0.95 },
{ "type": "IBAN",
"text": "DE89...",
"score": 0.99 }
],
"certificate_id": "cert_a1b2c3"
}
curl -X POST https://ki-shield.de/v1/redact \
-H "X-API-Key: ks_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"text": "Herr Dr. Müller (Tel. 0171-1234567) wohnt in Berlin.",
"mode": "redact"
}'
# => "Herr Dr. [PERSON] (Tel. [PHONE]) wohnt in [LOCATION]."
import
{ KiShield }
from
'ki-shield-sdk';
const shield = new KiShield({
apiKey: 'ks_live_YOUR_KEY'
});
const result = await shield.redact({
text: 'Max Mustermann, IBAN DE89...'
});
console.log(result.redacted);
// => "[PERSON], IBAN [IBAN]"
console.log(result.certificate_url);
// => "https://ki-shield.de/verify-page/cert_..."
Legal Compliance
Legally on the Safe Side
Built for the strictest data protection requirements in Europe.
GDPR Art. 25 – Privacy by Design
Personal data is pseudonymized before leaving the EU jurisdiction. No data transfers to third countries without legal basis.
EU AI Act – Transparency Requirement
Complete documentation of every AI interaction. Traceable who sent what data to which AI provider and when.
DE
German Servers (Hetzner, Nuremberg)
Processing exclusively in Germany. No US CLOUD Act, no FISA 702, no access by foreign authorities.
AES-256-GCM & Hybrid Signatures
All stored data AES-256-GCM Encrypted (Authenticated Encryption). Audit-Log mit hybriden Ed25519 + ML-DSA-65 Signaturen (post-quantum-sicher) und Hash-Chain für manipulationssichere Dokumentation.
Our USP
Zero Knowledge – two modes, one goal: your data stays yours
Server-ZK: All stored data is encrypted with your password (Argon2id + AES-256-GCM). No re-identification possible if database is accessed. Browser-ZK (NEU): PII detection and encryption run entirely in your browser – the server never sees plaintext, not even in memory.
zero-knowledge ~ architecture
# Your password generates the key
Password + Salt → Argon2id → 256-bit DEK
# Mode 1: Server ZK (default)
✓ Your browser: yes (session)
✓ RAM (Redis): temporary, volatile
✗ Database/Disk: never
# Mode 2: Browser ZK (NEW)
✓ Your browser: PII detection + encryption
✗ Server RAM: sees only pseudonyms
✗ Datenbank: ciphertexts only
✗ Operator: no access, ever
# What an attacker sees in the DB
Message: gcm$V4cPxqW29k3nR0Z8tzK7vHE2qF1mNpR5...
Title: gcm$QoFXjPwTmR9h2YLzNcK8vHE3qB1nMpR5...
Without password: ✗ all unreadable
Password-Based Key (Argon2id)
From your password, a 256-bit encryption key is derived via Argon2id – the strongest password hashing algorithm. This key exists only in volatile memory.
Operator Cannot View Stored Data
Without your password, no one can decrypt the stored data – not us, not an attacker, not a government agency. Note: During active processing, data briefly exists in plaintext in volatile memory — without storage, logging or disclosure. Processing exclusively on German servers.
Recovery Key as Safety Net
In your account settings, you can create a recovery key at any time. If you forget your password, use it to restore your data. Without recovery key and password: data is irretrievably lost.
Everything Encrypted – AES-256-GCM
Chat messages, conversation titles, API keys, audit logs – everything is AES-256-GCM encrypted with your personal key. Each user has their own key.
New
Browser ZK Mode – True Zero Knowledge
In Browser ZK mode, your plaintext data never leaves your browser. PII detection, pseudonymization and AES-256-GCM encryption run entirely locally. The server only sees pseudonymized data – at no point, not even in RAM.
Local PII Detection
25,000+ German first and last names, organization dictionary (DAX, banks, government agencies), IBAN, email, phone, health data – all rule-based detection in the browser.
Browser Encryption
AES-256-GCM via WebCrypto API. Key is derived from your ZK password via PBKDF2. Pseudonym mappings are stored encrypted on the server – recoverable only with your password.
BYOK – Your Key
Exclusively your own API key (BYOK). No server key, no fallback. Full cost control and data sovereignty – the server is just a pseudonymized proxy.
All plans (incl. Free)
No installation
One click in settings
Other Providers
- Operator can read all data
- Database theft = data compromised
- Government request = disclosure possible
- "Trust us" as security model
KI-Shield Zero-Knowledge Storage
- Zero-knowledge storage – operator cannot view stored data
- Database theft = worthless ciphertexts
- Government request = stored data unreadable
- Cryptography instead of trust
- Browser ZK: Server never sees plaintext NEW
Future-Proof
Post-Quantum Cryptography – ready for tomorrow
Quantum computers will be able to break classical encryption. KI-Shield already protects your audit chain today with hybrid signatures – both classical and quantum-safe.
audit-chain ~ hybrid signature
# Every audit entry is dual-signed
# 1. Classical signature (secure today)
Ed25519: 7b2e9f4a...c8d1
✓ Key: 256 bit
# 2. Post-quantum signature (secure tomorrow)
ML-DSA-65: a3f8b2c1...e7f9
✓ Standard: NIST FIPS 204
# Hash chain (tamper-proof)
SHA-256: prev_hash → content → next
✓ Hybrid signature verified
ML-DSA-65 (FIPS 204)
The NIST-standardized post-quantum algorithm (formerly CRYSTALS-Dilithium). Resistant to attacks from quantum computers – including future ones.
Hybrid Signatures – Double Security
Every audit entry is dual-signed: Ed25519 (classical, proven) and ML-DSA-65 (post-quantum). Even if one algorithm is broken, the other protects.
Tamper-Proof Audit Chain
SHA-256 hash chain links every entry to the previous one. Tampering with a single entry breaks the entire chain – immediately detectable, even after years.
EU Regulation Compliant
The BSI and the EU recommend transitioning to post-quantum cryptography. With KI-Shield, you are already compliant – before it becomes mandatory.
Why Post-Quantum Now?
Attackers can store encrypted data today and decrypt it later with quantum computers (Harvest Now, Decrypt Later). Audit logs often need to remain verifiable for 10+ years. That is why KI-Shield already signs quantum-safe today – so your compliance records are still valid in 2035.
Market Analysis
Trust-as-a-Service – the Verification Economy
Verifiable trust is becoming the scarcest resource in the digital world. KI-Shield provides the cryptographic infrastructure that enterprises need to use AI in a demonstrably secure way.
26,8
Mrd. USD Verifizierungsmarkt bis 2031
Gartner / MarketsandMarkets
169
B USD sovereign cloud by 2028
Gartner IaaS Prognose
57%
expect turbulent global conditions by 2028
World Economic Forum
42,8%
CAGR deepfake detection
Mordor Intelligence
Why Trust Is Becoming Infrastructure
The World Economic Forum identifies misinformation and disinformation as the most dominant risk over the next two years. AI-generated deepfakes, synthetic evidence and AI poisoning are destroying trust in digital information.
At the same time, cloud fragmentation is forcing companies to split their AI architectures across sovereign zones. 60% of multinational companies must geopatriate their AI infrastructure by 2028 (IDC).
Gartner warns: Companies that fail to invest in digital provenance by 2029 risk compliance damages in the billions.
Read the Full Market Analysis
16-page analysis with data from WEF, CFR, McKinsey, Gartner and IDC – the foundation for KI-Shield.
Macroeconomic Projections to 2028 →
trust ~ ki-shield stack
$ ki-shield --trust-stack
# Layer 1: PII Protection
42 Kategorien detected & pseudonymized
✓ Zero-Knowledge-Speicherung – we never see plaintext data
✓ Browser-ZK – plaintext never leaves the browser
# Layer 2: Cryptographic Proof
Ed25519 + ML-DSA-65 Hybrid Signature
SHA-256 Hash-Chain (manipulationssicher)
# Layer 3: Blockchain Anchoring
Polygon PoS – daily anchor
✓ Publicly verifiable, immutable
# Layer 4: Sovereignty
100% German infrastructure
No US cloud. No data transfer. GDPR-native.
# Layer 5: Audit Compliance
WORM-Archive + eIDAS-Zeitstempel
✓ Probative for regulatory authorities
→ All 5 protection layers active
Patent Filed
System and method for pseudonymized AI usage with zero-knowledge architecture
The 5 Capital-Critical Sectors by 2028
Energy & Cooling
$6.7T by 2030
Verification Economy
KI-Shield
Sovereign Cloud
$169B by 2028
Critical Infrastructure
$106T by 2040
Nearshoring
Supply chain autonomy
Transparency
What KI-Shield Can – and What It Cannot
We communicate our limitations openly. Only those who acknowledge their limits deserve trust.
What KI-Shield Does
- Automatic pseudonymization of 42 PII categories using NER, regex, keyword and context analysis
- Additional technical safeguard per GDPR Art. 25/32
- Encryption of all stored data (AES-256-GCM)
- Tamper-proof audit log with cryptographic signatures
- Manual tagging for data not automatically detected
What You Should Know
- Stylometric features (writing style, sentence patterns) and purely combinatorial re-identification are not automatically detected – for this we offer manual tagging as a supplement
- Data is processed exclusively in volatile memory on German servers – without storage or logging of plaintext
- KI-Shield is a technical safeguard (TOM) per GDPR Art. 25/32 – as part of your data protection concept
Data Processing Agreement (DPA)
As a data processor, KI-Shield provides a DPA in accordance with GDPR Art. 28. The DPA is automatically made available upon account creation.
KI-Shield complements your existing compliance strategy as an additional technical and organizational measure (TOM) – it does not replace a DPA with your AI provider.
Frequently Asked Questions
Answers for Decision Makers
The most important questions about data protection, technology and costs – answered clearly.
Without additional safeguards, using ChatGPT in law firms is problematic, as personal client data is transmitted to servers in the USA. KI-Shield reduces this risk through automatic pseudonymization: all identifiable personal data is replaced with placeholders before being sent to the AI.
Patient data is protected by medical confidentiality and the GDPR. Without pseudonymization, no patient data may be transmitted to AI providers. KI-Shield automatically pseudonymizes all sensitive data – patient names, diagnoses, insurance numbers – before it leaves your practice.
Yes, KI-Shield operates on the BYOK principle (Bring Your Own Key). You add your own API key. Advantage: Full cost control, no markups on token costs.
No. KI-Shield is API-compatible with over 10 AI providers. You simply change the API URL – instead of sending directly to OpenAI, requests go through KI-Shield. Integration in under 5 minutes.
Exclusively on German servers at Hetzner in Nuremberg. No US CLOUD Act, no FISA 702. Original data never leaves the KI-Shield server in the EU.
Der Free-Plan is permanently free (50 requests/month). Der Pro-Plan costs €99/month net. Der Business-Plan costs 349€/month with OpenRouter and Compliance reports. Der Enterprise-Plan starts at €1,999/month with RBAC API keys, PII-Redaction REST API, dedicated support, 99.9% SLA and on-premise option. On top, only your own AI costs with the provider (BYOK). No minimum contract period.
Yes. KI-Shield meets the transparency and documentation requirements of the EU AI Act. Every AI interaction is logged in the audit trail. The audit log is exportable (in Pro and Business plans).
KI-Shield offers two zero-knowledge modes: Server-ZK – the mapping table is encrypted with your password (Argon2id, 256-bit), all data stored AES-256-GCM encrypted. During processing, data briefly exists in server RAM (without storage or logging). Browser-ZK (NEW) – PII detection, pseudonymization and AES-256-GCM encryption run entirely in your browser. The server never sees plaintext – not even in RAM. Available for all plans, one click in settings.
42 PII categories in 4 groups: Core (names, email, phone, IBAN, tax ID, credit card, address, etc.), Technical (IP address, IMEI, MAC, GPS, driver license, etc.), GDPR Art. 9 (genetics, biometrics, ethnicity, politics, religion, union membership, sexuality) and Art. 10 + Life Areas (criminal law, children data, finances, employment, education, social benefits, insurance).
KISHIELDCAM — EXCLUSIVE FROM BUSINESS
KiShieldCam Feature
Auditor Access for External Reviewers
Give data protection officers, auditors or government agencies temporary read access to your evidence chain — without password sharing, without risk.
1
KiShieldCam — Auditor-Token
// Create token
Label:
"TÜV Audit Q1"
Valid:
7 days
✓ Token generated
aud_k7x9m2p4r8...
Create Token
One click in the app. Choose a label and validity (1–30 days). Revocable at any time.
2
Copy link
ki-shield.de/verify
?token=aud_k7x9m2...
?token=aud_k7x9m2...
Link copied!
E-Mail
Expires: March 27, 2026
Share Link
Copy and send via email. The reviewer needs no account and no installation.
3
Audit-Report (Read-Only)
Audit entries
247
Hash-Chain
✓ Intakt
Signaturen
✓ 247/247
Blockchain
✓ Verifiziert
Status
COMPLIANT
Reviewer Verifies
Full read access to the evidence chain. Signatures, hashes, blockchain — everything verifiable.
Zero Trust
No password sharing. Token-based.
Time-Limited
1–30 days. Automatically expires after.
Instantly Revocable
One click — access immediately revoked.
Read-Only
Read only. No writing, no deleting.
Teil von KiShieldCam — court-proof evidence photos & videos
More about KiShieldCam Who We Are
The Team Behind KI-Shield
No startup theater. Two people, one product.
team ~ ki-shield
$ whoami --team
JOHANNA
Role: UX Design · Social Media · Branding
Focus: User experience, visual design, communication
Status: ████████████ active
RENÉ
Role: Datenbank · Backend · AI Automation
Focus: Architecture, cryptography, PII pipeline
Status: ████████████ active
$ cat /etc/ki-shield/philosophy
> No VC. No hype. German engineering for European law.
Contact
Talk to Us
Questions about KI-Shield, compliance or integration? We will get back to you within 24 hours.
Message Sent
Thank you! We will get back to you as soon as possible.