Last updated: December 31, 2024

Short answer: Elasticbrains uses GLiNER as a local pre-processing layer to detect and permanently remove personal data (PII) before requests go to LLMs like GPT-4 or Claude. The AI is briefed to respond intelligently to detected PII - without re-substitution. This allows regulated industries like insurance, pharma, and legal to leverage full AI power in a GDPR-compliant manner.

The Problem: Why Many Companies Cannot Use AI

Many companies want to deploy AI solutions but face a fundamental problem: Their data contains personal information (PII) that cannot simply be transferred to US servers under GDPR.

GDPR Risk

Personal data may not be transferred to US servers without legal basis.

Industry Regulations

Pharma, finance, and insurance have strict compliance requirements.

Sensitive Documents

Contracts, claims, patient records - full of personal data.

Our Solution: Local PII Filtering + Intelligent AI Briefing

We developed an architecture that solves the problem at its root: GLiNER as a local pre-processing layer detects and removes personal data before it reaches external LLMs.

1

[EN] Input with PII

"I am John Smith from Main St. 5..."

2

[EN] GLiNER Filter (local)

[EN] PII detected and permanently removed

3

[EN] LLM (GPT/Claude)

[EN] Cleaned data + AI is briefed

The Key Difference: No Re-Substitution!

Unlike other solutions, we do not replace PII with placeholders that are re-substituted later. Instead:

  • PII is permanently removed - no mapping table as risk
  • The AI is briefed and responds intelligently to detected PII
  • PII never appears in output - true anonymization
How the AI responds:

User: "Hello, I am John Smith from Main St. 5..."

AI: "I see you just entered personal information - it's automatically anonymized by our system, so that's working well! But let's jump right into your actual concern: How can I help you today?"

Real-World Examples from Regulated Industries

Insurance Inbox

Claims contain names, addresses, policy numbers, and license plates. GLiNER detects and removes all PII locally. The AI then categorizes and analyzes only the cleaned case facts.

[EN] Result: Automated claims processing without GDPR risk

Legal Document Analysis

Complaints, responses, and expert opinions are analyzed. GLiNER detects party names, case numbers, and addresses. The AI creates structured summaries - without personal data.

[EN] Result: Significant time savings in case analysis

Voice AI / Phone Assistant

Callers mention names and customer numbers. GLiNER detects PII in real-time (Speech-to-Text → Filter → LLM). The assistant confirms anonymization and continues the conversation professionally.

[EN] Result: 24/7 availability without data protection concerns

Pharma & Healthcare

Patient data, study results, and medical reports are processed. GLiNER detects patient names, birth dates, and hospital IDs. AI experts can work on cleaned data.

[EN] Result: Compliance-compliant AI usage in healthcare

Why This Is a True USP

Aspect
Other Providers
Elasticbrains
PII Handling
Direct API calls or re-substitution
Local filter + permanent cleaning
Mapping Table
Yes - additional risk
No - no mapping needed
PII in Output
Possible (after re-substitution)
Never - true anonymization
Regulated Industries
"We use Azure OpenAI"
Any LLM GDPR-compliant

Our Technology Stack

We combine various technologies into a robust data protection architecture:

  • GLiNER

    Open-source NER model for local PII detection. Detects names, addresses, phone numbers, emails, IBANs, and more.

  • n8n Workflow Orchestration

    Visual orchestration of AI pipelines. GLiNER node → LLM node → Output in a traceable workflow.

  • LLM Integration (flexible)

    GPT-4, Claude, Azure OpenAI, Gemini, or Mistral - the local filter enables GDPR-compliant use of any model.

  • Intelligent System Prompting

    The AI is briefed to handle detected PII situations professionally and continue the conversation seamlessly.

Frequently Asked Questions

Is GLiNER really 100% local?

Yes, GLiNER runs completely on your server or in your cloud environment. No data leaves your infrastructure for PII detection.

Which LLMs can I use then?

All of them! Since PII is removed before the LLM call, you can use GPT-4, Claude, Gemini, Azure OpenAI, or Mistral in a GDPR-compliant manner.

What happens with false positives?

GLiNER is optimized for precision. Additionally, industry-specific rules can be defined to minimize false detections.

How fast is the processing?

GLiNER is a lightweight model with minimal latency. PII detection typically takes only milliseconds per request.

Ready for GDPR-Compliant AI?

Let's design your AI strategy in a data protection-compliant way together.